Emotion Response Patterns to Transient Stimuli in Migraineurs and Non-Migraineurs Modulation of Affective States As a Step Towards Non-Invasive Treatment of Migraine

UNIVERSIDADE DE LISBOA

FACULDADE DE CIÊNCIAS FACULDADE DE LETRAS FACULDADE DE MEDICINA FACULDADE DE PSICOLOGIA

EMOTION RESPONSE PATTERNS TO TRANSIENT STIMULI IN MIGRAINEURS AND NON-MIGRAINEURS MODULATION OF AFFECTIVE STATES AS A STEP TOWARDS NON-INVASIVE TREATMENT OF MIGRAINE

Ana Pesquita

MASTER IN COGNITIVE SCIENCE

2010

UNIVERSIDADE DE LISBOA FACULDADE DE CIÊNCIAS FACULDADE DE LETRAS FACULDADE DE MEDICINA FACULDADE DE PSICOLOGIA

EMOTION RESPONSE PATTERNS TO TRANSIENT STIMULI IN MIGRAINEURS AND NON-MIGRAINEURS MODULATION OF AFFECTIVE STATES AS A STEP TOWARDS NON-INVASIVE TREATMENT OF MIGRAINE

Ana Pesquita Thesis supervised by Prof. A.M. Sebastião, PhD co-supervised by Prof. Dr. P.F.M.J. Verschure

MASTER IN COGNITIVE SCIENCE

2010

ACKNOWLEDGMENTS

During the making of this thesis I have encountered many people who contributed in different ways to the work and to the enrichment of the personal experience that it constituted. For their supervision, I want to thank Ana Sebastião and Paul F.M.J. Verschure. For creating a space for discussions on Cognitive Science inside Lisbon University, I want to thank the Scientific Committee of the Master in Cognitive Sci- ence and all the remarkable folks I encountered during these studies. The initial phase of this work was developed under the financial support of the INOV-Art Scholarship, an initiative of the Portuguese Ministry Of Culture. Synthetic Perceptive Emotion Cognition Systems Laboratory (SPECS), at Pompeu Fabra University, supported the development of this work and contributed deeply to the scientific insights in it expressed. Aleksander Valjamae supervised the initial phase of this master thesis and, together with Paul F.M.J. Verschure, is the author of the original idea of the AMDB system. Arnau Espinosa and Eliza-Nefeli Tsaoussi collaborated on software development and testing. My thanks also go to Patricia Pozo-Rosich and Joan Durà from the Hospital Vall D’Hebron, and to the Nexes association for their tireless efforts in the search for participants. Special thanks go to all the women who participated in the empirical study, without whom the development of this work would not have been possible. I seize this moment to express my huge gratitude to my family, a clan of wise thinkers and brave doers. Finally, my last thank you is saved for Ulysses Bernardet, the author of the first implemented version of the AMDB software, that was the starting point of my contribution to this SPECS project. Ulysses’ energetic mind and sincere heart balanced their presence in every step of this journey, and for that I feel extremely grateful.

ABSTRACT

This thesis was developed in the context of the scientific study of emotion, and focuses on the modulation of affective states through perceptual stimuli. Two goals were pursued during the course of this master project. Firstly, the development of the Affective Multimodal Data Base (AMDB) system, a tool for multimodal emotion induction, and secondly, to conduct an empirical study, using the AMDB, that aimed at probing the modulation of affective response patterns in migraineurs. The AMDB system is a software platform for generating affective stimuli by combining sounds, pictures and videos, available in normalized affective stimuli libraries, into multimodal stimuli sequences to be used in emotion induction scenarios. Con- tributions of this master project to the development of the AMDB system include the addition of a graphical-user-interface, implementation of synchronized stimuli presentation and physiology response recording, improvement in the modular implementation of the system, and finally, realization of the first test-case of the AMDB system. It was decided to focus the empirical study on migraine, because it is a widespread pathology that presents a clear link with emotion mechanisms. The empirical study aimed at investigating dynamic patterns of affective responsiveness and modulation in migraineurs with aura. The findings from this study suggest an affective profile specific to migraineurs with aura characterized by an enhanced impact of unpleasant stimuli, potentiation of negative affective states when repeatedly exposed to non-pleasant stimuli, and high susceptibility to disrupt positive affective states in the presence of unpleasant and neutral environmental stimuli. This study hopes to contribute to the understanding of affective response patterns specific to migraine, and to provide insights for the design of therapy approaches to migraine based on emotion modulation, that could in the future help migraine sufferers ease the disease burden through the use of accessible media technologies.

Keywords: emotion modulation, affective proﬁle, migraine, pain, non-invasive therapy

RESUMO

Esta dissertação foi desenvolvida no contexto do estudo científico da emoção, e foca a modulação de estados afectivos através de estímulos perceptuais. O progresso deste trabalho foi guiado por dois objectivos principais: primeiro, o desenvolvimento do sistema Affective Multimodal Data Base (AMDB), uma ferramenta para indução de estados afectivos; e segundo, a realização de um estudo empírico, utilizando o AMDB, com o objectivo de examinar a modulação de padrões de respostas afectivas em pessoas que sofrem de enxaqueca com aura. O sistema AMDB é uma plataforma de software para a geração de estímulos afectivos através da combinação de sons, imagens estáticas e vídeos a ser utilizados em contextos de indução de estados afectivos. Contribuições deste mestrado para o desenvolvimento do AMDB incluem a adição de um interface gráfico para o utilizador, a implementação de apresentação de estímulos e gravação de respostas fisiológicas de forma sincronizada, melhoramentos na implementação modular do sistema, e por fim, a realização de um primeiro caso de teste do AMDB. A decisão de focar o estudo empírico, desenvolvido dentro do âmbito deste trabalho, na patologia da enxaqueca foi motivada pelo facto desta ser muito difundida e apresentar relações claras com os mecanismos de emoção. Os resultados deste estudo sugerem um perfil afectivo específico às pessoas que sofrem de enxaqueca com aura caracterizado por susceptibilidade intensificada a estímulos desagradáveis, po- tenciação de estados afectivos negativos em situações de repetida exposição a estímu- los neutros ou desagradáveis, e disposição aumentada para a interrupção de estados afectivos positivos na presença de estímulos ambientais de cariz emocional neutro ou desagradável. Este estudo espera contribuir para a compreensão dos padrões de respostas afectivas específicos na patologia da enxaqueca, servindo como ponto de partida para o design de abordagens terapêuticas baseadas na modulação de estados afectivos.

Palavras-chave: modulação de respostas emocionais, perﬁl afectivo, enxaqueca, dor, terapias não-invasivas

SUMÁRIO

O estudo da emoção foi abraçado pela neurociência e ciência cognitiva apenas recentemente. Os enigmas da emoção, uma função complexa do cérebro intimamente ligada à experiência subjectiva e ao comportamento, foram durante um longo período de tempo deixados fora do cerne científico. No século passado, a investigação da emo- ção proliferou inspirada pela busca de conhecimento acerca de como a emoção modula o comportamento e funções cognitivas como a atenção, a memória e a tomada de de- cisões. Tendências na neurociência focaram-se nos fundamentos neurofisiológicos da emoção, e é agora aceite que a emoção é um importante interveniente para a compre- ensão holística da mente e do cérebro. É neste contexto da ciência da emoção que a presente dissertação foi desenvolvida. O objectivo central deste trabalho foi investigar a modulação de padrões de respostas afectivas em pessoas que sofrem de enxaqueca. A decisão de focar a enxaqueca, uma patologia que envolve episódios severos de dor de cabeça, foi motivada pelo facto da enxaqueca ser uma condição clínica que apresenta relações claras com os mecanismos de emoção. A enxaqueca é também um tópico de interesse dado ser um problema globalmente pouco reconhecido e pouco tratado. A Organização Mundial de Saúde (OMS) anunciou números relevantes que colocam a enxaqueca entre as maiores causas de incapacidade. Recentemente, o papel da emoção na modulação da dor tem recebido especial aten- ção devido a vários estudos que, por um lado suportam o importante contributo da elaboração cognitiva para o desenvolvimento de estratégias por parte dos pacientes para lidar com a dor, e que por outro lado sugerem que mecanismos neurofisiológicos ligados ao processamento emocional influenciam a modulação de nocicepção. Conse- quentemente, a compreensão da dimensão da emoção nas desordens de dor crónica está actualmente no cerne do desenvolvimento de abordagens terapêuticas não-invasivas, que pretendem contribuir para o alívio do peso da dor nas vidas de pessoas que sofrem deste tipo de condições clínicas. O estudo empírico desenvolvido no âmbito deste trabalho teve como objectivo examinar as respostas a estimulação afectiva de pessoas que sofrem de enxaqueca. Em particular, foi estudada a flexibilidade na adaptação das resposta afectivas a mudanças no contexto da estimulação afectiva. No sentido de abordar o objectivo de investigação proposto desenvolvi duas tarefas em separado: 1. Colaboração no desenvolvimento de uma ferramenta de software para modula- ção de estados afectivos utilizando estímulos multi-modais, o sistema Affective Multimodal Data Base (AMDB) 2. Realização de um estudo empírico, utilizando o sistema AMDB, junto de uma Sumário VIII

amostra da população de pessoas que sofrem de enxaqueca com aura e participantes de controlo O sistema AMDB tem como objectivo tornar disponível uma ferramenta integrada para a apresentação em tempo real de sequências de estímulos afectivos multi-modais, i.e. através da combinação de sons, imagens estáticas e vídeos. As funcionalidades centrais oferecidas pelo sistema AMDB incluem o acesso central a várias bibliotecas normalizadas de estímulos afectivos, combinação modular de estímulos sonoros e visuais em sequências, integração de apresentação de estímulos e gravação de respostas fisiológicas de forma sincronizada, e acesso estruturado a dados experimentais. O estudo empírico foi conduzido junto de uma amostra de dezoito mulheres que sofrem de enxaqueca com aura e vinte e duas participantes de controlo, com o objectivo de investigar diferenças entre os perfis afectivos de ambos os grupos. O estudo foi realizado utilizando imagens estáticas como estímulos afectivos, na sequência de resultados promissores relativos à atenuação da dor via indução de emoções positivas obtida através de estímulos visuais. Os dados experimentais foram recolhidos através de julgamentos subjectivos e gravação de respostas fisiológicas. Fez parte do âmbito deste projecto a análise dos dados recolhidos através de julgamentos subjectivos. As contribuições desta dissertação de mestrado estão também distribuídas em duas dimensões. Por um lado, as contribuições para o desenvolvimento do AMDB incluem: 1. Adição de um interface gráfico para o utilizador 2. Implementação de apresentação de estímulos e gravação de respostas fisiológi- cas de forma sincronizada 3. Melhoramentos na implementação modular do sistema 4. Realização de um primeiro caso de teste do AMDB O software AMDB desenvolvido pode ser utilizado no estudo de perfis afectivos em diversos cenários, não estando limitado ao uso específico no estudo empírico desenvolvido no âmbito deste trabalho. Por outro lado, um estudo de investigação focando os padrões de respostas afectivas de pessoas que sofrem de enxaqueca com aura. Os resultados do estudo empírico sugerem um perfil afectivo específico na patologia da enxaqueca com aura caracterizado por: 1. Susceptibilidade intensificada a estímulos desagradáveis 2. Potenciação de estados afectivos negativos em situações de repetida exposição a estímulos neutros ou desagradáveis 3. Disposição aumentada para a interrupção de estados afectivos positivos na pre- sença de estímulos ambientais de cariz emocional neutro ou desagradável Estes resultados são consistentes com a associação da enxaqueca com disfunções da emoção expressas pela comodidade entre enxaqueca e desordens de depressão e an- siedade, a predominância de traços de personalidade de neuroticismo na população de pessoas que sofrem de enxaqueca, e por último, o proeminente papel do stress no desencadeamento de crises de enxaqueca. A relação entre a enxaqueca e os mecanismos de emoção faz desta patologia uma Sumário IX candidata ideal para o desenvolvimento de terapias não-invasivas baseadas na modu- lação de estados afectivos. Neste contexto, a estimulação afectiva poderá ser induzida através de um fluxo contínuo de sons, vídeos e imagens estáticas à semelhança de um filme cuja a edição é guiada em tempo real pelo balanço entre estados afectivos a atingir, que têm implicações na modulação da dor, e os estados afectivos reais dos pacientes, estimados através da análise automatizada de respostas subjectivas e dados fisiológicos. Desenvolvimentos na área dos média, dedicada à criação de narrativas in- teractivas e motores de jogos, oferecem soluções que podem ser adaptadas ao contexto clínico, no entanto, é necessário aprofundar o conhecimento acerca do perfil afectivo dos pacientes de enxaqueca. O sistema AMDB apresenta um primeiro protótipo para a geração contínua de es- timulação afectiva multi-modal. O desafio de criar padrões de estimulação emocional contínua poderá ser investigado numa área de interface entre a ciência e a arte. Técni- cas de narrativa têm vindo a ser desenvolvidas e testadas em vários domínios da arte ao longo da história da cultura humana. Um aspecto central destas técnicas refere-se ao desenvolvimento temporal da apresentação de circunstâncias de forma dinâmica, que resulta numa componente de envolvimento emocional por parte dos destinatários. Consequentemente é possível antecipar que a construção de narrativas específicas para uso na modulação emocional da dor terá de ser baseada no conhecimento dos padrões de respostas afectivas dos pacientes a dinâmicas no conteúdo emocional da narrativa. Este estudo espera contribuir para a compreensão dos padrões de respostas afectivas específicos à enxaqueca, servindo como ponto de partida para o design de abordagens terapêuticas baseadas em modulação de estados emocionais. Estas novas abordagens visam contribuir para o alívio do peso que a enxaqueca traz à vida das pessoas que dela sofrem, através da utilização de tecnologias média de grande acessibilidade.

CONTENTS

1. Introduction ...... 1 1.1 Migraine pathology ...... 2 1.1.1 Epidemiology and clinical features ...... 2 1.1.2 Comorbidity with depression and anxiety disorders ...... 4 1.2 Pathophysiology of migraine ...... 8 1.2.1 Mechanisms involved in migraine attacks ...... 8 1.2.2 Common neural pathways in migraine and depression . . . . 16 1.3 Migraine treatment approaches ...... 19 1.3.1 Available approaches ...... 19 1.3.2 Emotion modulaion ...... 21

2. Methods ...... 29 2.1 Experimental design ...... 29 2.1.1 Temporal structure of the phases ...... 30 2.2 Stimuli ...... 30 2.3 Sample ...... 32 2.4 Procedure ...... 33 2.5 Apparatus ...... 34 2.5.1 Development of the Affective Multimodal Data Base (AMDB) system ...... 34

3. Results ...... 39 3.1 Affective responses to single pictures ...... 39 3.1.1 Validation of the affective pictures classiﬁcation system . . . 39 3.1.2 Assessment of IAPS affective induction predictions ...... 41 3.1.3 Comparison between migraineurs and non-migraineurs responses 41 3.1.4 Analysis of the effects in repeated exposure to affective stimuli 44 3.2 Affective responses to transient stimuli ...... 45 3.2.1 Analysis of the Valence Inhibition effect ...... 46 3.2.2 Probing the temporal modulation of Valence Inhibition and arousal modulation ...... 50

4. Discussion and Conclusion ...... 55 4.1 Summary of thesis work ...... 55 Contents XII

4.2 Summary of results ...... 55 4.2.1 Single stimuli ...... 56 4.2.2 Transient stimuli ...... 56 4.3 Contextualization of results ...... 57 4.4 Critical evaluation of the experimental paradigm ...... 58 4.5 Future directions ...... 59

List of Figures ...... 65

List of Tables ...... 66

References ...... 67

Appendix 75

A. Treatment approaches followed by MA participants ...... 76

B. Legal consent ...... 77

C. Behavioral Inhibition Behavioral Activation questionnaire (BISBAS) .... 78 1. INTRODUCTION

The study of emotion was embraced by neuroscience and cognitive science only recently. The riddles of emotion, a complex brain function tightly linked to subjective experience and behavior, were for a long time left out of the scientific mainstream. In the last century, emotion research arouse in the quest for understanding how emotion modulates behavior and high cognitive functions such as attention, memory and decision-making. Trends in neuroscience focus on the neurophysiological underpin- nings of emotion, and emotion is now accepted to play an important role in a holistic understanding of mind and brain. It is in this context of emotion science that the thesis presented here was developed. The main goal of this work was to investigate the modulation of affective response patterns in migraineurs. It was decided to focus the empirical study on migraine, a pathology involving episodes of severe head pain, because it is a widespread clinical condition that presents a clear link with emotion mechanisms. Migraine link with emotion mechanisms is revealed by migraine comorbidity with depression and anxiety disorders, the prominent role of distress as a trigger to migraine attacks, and actual mood abnormalities experienced during migraine attacks. The study conducted aimed at examining migraineurs responses to affective stimulation, and in particular, migraineurs response flexibility in adaption to shifts of emotion stimulation. Migraine is of further interest because it is an under-recognized and under-treated global problem. The World Health Organiza- tion (WHO) announced striking numbers putting migraine among the biggest causes of disability. Recent attention has been given to the role of emotion in pain modulation and several studies support the role of cognitive elaboration in patients skills in coping with pain, suggesting that neurophysiological mechanisms linked to emotion influence nociceptive modulation. Hence, furthering the understanding of the emotion dimension in chronic pain disorders is currently in the core of the development of non-pharmacological approaches that may help sufferers to lower the burden caused by pain. To fulfill the goal of investigating the modulation of affective response patterns in migraineurs, I conducted two separate tasks: 1. The development of a software tool for multimodal emotion induction, the Af- fective Multimodal Data Base (AMDB) system; 2. The realization of an empirical study using AMDB with a group of migraineurs with aura and a control group The goal of the AMDB is to provide an integrated tool for real-time presentation of multimodal stimuli sequences, i.e. by combining affective sounds, pictures and 1. Introduction 2 videos. The main functionalities offered by the AMDB system are centralized access to various libraries of normalized affective stimuli, modular combination of the audio and visual stimuli into sequences, integrated physiology recording synchronized with stimuli presentation and structured storage and access to experimental data. The empirical study was conducted with a group of migraineurs with aura and a group of control subjects, aiming at investigating differences between their affective profiles using visual stimuli, following promising results of recent studies on pain attenuation via positive emotion induction achieved by visual stimuli. The contributions of this thesis are then two-fold as well: on one hand the development of a tool for multimodal emotion induction which can be used to assess affective profiles in multiple scenarios, not limited to the one used for the empirical study, and on the other hand, an investigation on the affective profiles of migraineurs with aura. The remainder of this thesis is organized as follows: continuing on chapter 1 I will give an overview of the state of the art and theoretical background pertinent to this thesis; chapter 2 describes the methodology employed; chapter 3 presents the analysis of the results obtained in the empirical study and finally chapter 4 is dedicated to discussing these results and drawing a conclusion, highlighting possible avenues of future work.

1.1 Migraine pathology

In this section I will start by brieﬂy presenting migraine epidemiology data to stress the widespread occurrence of the disease and provide a demographic view on the disease burden. Next, I will address the heterogeneous constellation of clinical symptoms experienced by migraineurs. Afterward, I will draw near the speciﬁc focus of this manuscript by dedicating attention to migraine’s comorbidity with anxiety and depression disorders.

1.1.1 Epidemiology and clinical features Migraine, a pathology with an heterogeneous symptom constellation that includes episodes of severe head pain, is a global problem of great impact on both the patient and society. Recent World Health Organization (WHO) campaigns directed at understanding the scale and scope of migraine epidemics, revealed that migraine worldwide prevalence is 12% (16% in women and 8% in men). Migraine is most prevalent among adults, and most common in Europe (15%) and reportedly less common in Africa (5%). Migraine is positioned at the highest level of neurological disability disorders (i.e. ranging from 0.7 to 1.0, on a scale from 0 to 1) according to the WHO criteria of disability-adjusted life years (DALY), computed by multiplying the sum of years of life lost (YLL) due to premature mortality with years lived with disability (YLD). Fur- thermore, migraine ranks 19th among all causes of disability (12th in women) based on the years-lived-with-disability criteria (deﬁned as incidence multiplied by disability 1. Introduction 3

Repeated episodic headache (4 to 72 hours) with the following features: Any two of: • Unilateral • Throbbing • Worsened by movement • Moderate or severe Any one of: • Nausea/vomiting • Photophobia and phonophobia

Fig. 1.1: International Headache Society features of migraine (from Goadsby, 2009) multiplied by disease burden). These striking numbers highlight the global effects of a common pathology and are in the motive of the development of the “Lifting the Bur- den project”, an international campaign committed to reduce the burden of headache worldwide (Moskowitz & Buzzi, 2010). Migraine is a complex group of disorders with several phenotypes and heterogeneous clinical symptoms that vary widely between patients and even within the same patient. The International Headache Society (IHS) who defines diagnostic criteria for headache disorders, organized migraine phenotypes into 19 different migraine subtypes (Olesen & Steiner, 2004). In common migraine subtypes, diagnostically important features usually include unilateral or bilateral head pain often with a throbbing quality that is worsened by movement, photophobia, phonophobia and/or osmopho- bia and autonomic related symptoms such as nausea and vomiting (Fig. 1.1). Proper diagnosis is essential to ensure the choice of an adequate treatment approach. Atten- tion to family history is relevant in the diagnosis procedure given the strong genetic component of migraine – up to 50% (Pietrobon & Striessnig, 2003). Migraine crisis usually follow a sequential pattern categorized by four stages, i.e. prodrome, aura, head pain, and resolution or a postdrome. Nonetheless, often patients do not experience all of the stages described (e.g. migraine without aura). In the prodrome phase migraineurs experience premonitory symptoms of the migraine attack. These include yawning, malaise, acute mood changes, food cravings, nausea, as well as clinical signs of sensory hyper-excitability, i.e. photophobia, phonophobia, hyperosmia or cutaneous allodynia. Patients are often capable of predicting an impending migraine attack based on the premonitory symptoms. A study using an electronic diary system showed that most patients were able to predict 72% of the attacks based on symptoms they believed to represent the prodromal phase of migraine (Giffin et al., 2003). It is in the premonitory phase that migraine triggers have a prominent role in promoting the development of the migraine crisis. Some evidence points to migraine attacks being triggered by environmental factors (e.g. weather, lights, noise, and odors), nutritional factors (e.g. alcohol intake, caffeine withdrawal, skip- 1. Introduction 4 ping meals, and possibly dehydration), psychological factors (e.g. distress, sleeping problems, fatigue, and tiredness), and in female migraineurs cyclic hormonal changes (Levy, Strassman, & Burnstein, 2009; Wober & Wober-Bingol, 2010). Migraine aura can take the form of visual symptoms, paresthesias1, or speech disturbances. Often migraine patients describe visual aura as distorted vision starting at the periphery and extending to fully involve the entire visual fields. In most of the occurrences an aura begins and ends before the migraine headache, lasting up to 60 minutes. Not all migraineurs experience aura symptoms, and some patients who have frequent attacks with aura also experience attacks without aura. The IHC classifies migraine with aura (MA) and migraine-without-aura (MO) into distinct migraine subtypes (Benoit, 2009). Migraine headache is a severe pain, often characterized by a throbbing unilateral or bilateral sensation, worsened by physical activity. The pain lasts between 4 to 73 hours. For some migraineurs, pain is accompanied by mood distortions (e.g. increased anxious state), autonomic dysfunction symptoms (e.g. nausea, vomit, piloerection), and sensory sensitivity (e.g. cutaneous allodynia) (Levy et al., 2009; Benoit, 2009; Moskowitz & Buzzi, 2010). The pain described by migraineurs is similar to the pain reported in other headache syndromes, but the above described symptom constellation helps clarify the diagnosis. The finalizing phase of a migraine crisis is the postdrome, characterized by the persistence of symptoms beyond the resolution of the headache. Some of these symptoms begin in the headache phase, whereas others only appear during the postdrome phase. During postdrome, patients often report loss of appetite, nausea, muscle tension, fatigue, and mental confusion. This phase can last between 1 to 2 days, extending the disability period produced by a migraine crisis (Cady, Schreiber, & Farmer, 2004).

1.1.2 Comorbidity with depression and anxiety disorders In the following, I will describe findings from studies focusing on the association of migraine with specific personality traits and relate these findings with the widely found comorbidity between migraine and psychiatric disorders of depression and anxiety. The clinical profile of migraineurs includes comorbid pathologies, as migraine appears significantly associated with the incidence of other disorders. The strongest associations with migraine are allergies, hypotension, hypertension, stroke, and depression and anxiety (Moskowitz & Buzzi, 2010). Literature focusing on the relationship between migraine and emotion disorders shows that migraine is consistently associated with diverse personality traits and psychiatric conditions. Evidence suggests a strong and systematic continuous link between personality and psychopathology. Hence, measuring personality traits may help researchers identify individual tendencies towards the development of specific psychopathologies (Corr & Matthews, 2009).

1 Paresthesia is a sensation of tingling, pricking, or numbness in the skin with no apparent long-term physical effect. 1. Introduction 5

Several studies addressing personality traits in migraine show that migraineurs tend to have distinctive personality characteristics (Table 1.1), i.e. neuroticism (K. Merikan- gas, 1993a; Fan, Gu, & Zhou, 1999), hostility (Bag, Hacihasanoglu, & Tufekci, 2005), depression and anxiety traits (Tan, Suganthi, Dhachayani, Rizal, & Raymond, 2007), high harm avoidance (Sánchez-Román et al., 2007; Abbate-Daga et al., 2007), high persistence, low self-directedness, difﬁcult anger management and tendency to hyper control (Abbate-Daga et al., 2007). Some of these traits may be closely related to psychiatric disorders of depression and anxiety found in co-occurrence with migraine (Corr & Matthews, 2009).

Tab. 1.1: Population based studies focusing on the relationship between personality traits and migraine. MO = Migraine without aura, MA = Migraine with Aura, CM = Chronic migraine, EM = Episodic migraine, TTH = Tension-Type Headache.

Reference Sample Personality As- Summary of results sessment K. Merikan- migraine vs. Freiburg Person- Migraineurs scored higher in neuroti- gas, 1993a no migraine ality Inventory cism scales than non-migraineurs Fan et al., migraine vs. Minnesota Multi- Migraineurs scored higher that non- 1999 no migraine phasic Personal- migraineurs in neuroticism scales and ity Inventory respective subsets (hipocondriasis, depression, hysteria), and in schizophrenia scales Karakurum CM vs. EM Minnesota Mul- CM showed higher scores of neuroti- et al., 2004 tiphasic Person- cism, depression, and social introver- ality Inventory sion than EM (MMPI) Bag et al., migraine vs. The Buss-Durkee Headache sufferers scored higher in the 2005 TTH vs. no Hostility Inven- hostility dimension than healthy sub- headache tory jects disorder Abbate- MO vs. no Temperament MO showed high harm avoidance, high Daga et al., migraine and Character persistence and low self-directedness in 2007 Inventory comparison to non-migraineurs Abbate- MO vs. no State-Trait Anger MO scored higher in difﬁcult anger Daga et al., migraine Expression management and tendency to hyper con- 2007 Inventory trol dimensions than non-migraineurs Tan et al., migraine vs. Minnesota Multi- Migraineurs scored higher in depres- 2007 no migraine phasic Personal- sion scales and anxiety scales than non- ity Inventory-2 migraineurs Sánchez- migraine vs. Temperament Migraineurs scored higher in harm Román et no migraine and Character avoidance scales than non-migraineurs al., 2007 Inventory

Population based studies show an association between migraine and psychiatric 1. Introduction 6 disorders, which is strongest for the case of depression and anxiety disorders (Ta- ble 1.2). Noticeably, migraineurs with aura present a higher risk of suffering from associated psychiatric disorders than migraineurs without aura. These psychiatric disorders include major depression disorder, bipolar disorder I, bipolar disorder II, panic attack disorder, obsessive-compulsive disorder, generalized anxiety disorder (as defined by the Diagnostic and Statistical Manual of Mental Disorders DSM-IV-TR Fourth Edition (Text Revision), 2000) (Torelli & D’Amico, 2004; Frediani & Villani, 2007). Additionally migraineurs show a high risk of development of chronic substance abuse (Radat & Swendsen, 2005). Understanding the etiological relationship between the association of migraine and psychiatric disorders is of great importance, since it may guide therapeutic approaches and provide ground for new theories on the concomitant biological and environmental subtract of migraine and specific psychopathologies. Following this research direction, studies examining the order of onset of migraine and psychiatric disorders showed that, on one hand, only phobic disorders appear to predict the onset of migraine, and on the other hand, migraine and depression or panic disorders don’t have a defined chronology. Migraine’s unclear causal link with depression and panic disorder gives rise to two hypothesis: either the causal link is symmetrical (each disorder is a risk factor of the other), or there is a strong genetic or environmental risk factor leading to the co-occurrence of both syndromes (Radat & Swendsen, 2005). These hypothesis can only be addressed through further understanding of the underlying mechanisms that lead to the comorbidity. Although no overlap exists in chromosomal locations of depression and migraine, as identified by linkage studies (Pietrobon & Striessnig, 2003), some preliminary results support the genetic basis of migraine and depression comorbidity. Both syndromes are therapeutically responsive to drugs that act on serotonin transmission, including selective serotonin re-uptake inhibitors and triptans. Genetic research investigating the relationship between migraine and depression focuses on the serotonergic system in the brain. The gene 5HTT is proposed as a potential shared genetic factor of both disorders, due to its expression on serotonin related mechanisms (Schur, Noonan, Buchwald, Goldberg, & Afari, 2009). Findings from population-based studies add some support to the genetic involvement in migraine and depression comorbidity. A recent study compared both, inheritance estimates between migraine with depression and migraine without depression, and inheritance scores of depression between migraineurs and controls. This investigation found a correlation between depression and migraine with aura, supporting the positive contribution of genetic factors to the comorbidity phenomena, particularly for the case of migraine with aura, also considered to have a stronger genetic component (Stam et al., 2010). In addition, results from twin studies investigating the occurrence of migraine and depression, and addressing the relative contributions of genetic and environmental factors, suggest that the association between both disorders may be due in part to shared genetic risk factors. Nevertheless the role of environmental factors in the etiology of depression and/or migraine was found to have a considerable influence (Schur et al., 2009). 1. Introduction 7

Tab. 1.2: Population based studies of migraine comorbidity with psychiatric disorders. Table adapted from Radat & Swendsen, 2005. DSM = diagnostic and statistic manual, DIS = diagnostic interview schedule, CIDI = composite international diagnostic interview, SPIKE = structured psychopathological interview and rating of the social consequences for epidemiology, TTH = tension-type headache.

Reference Sample Psychopathology Summary of results Assessment Breslau, migraine vs. no DSM-IIIR, DIS Signiﬁcant association between 1991 migraine migraine and anxiety disorders and substance abuse Breslau & migraine vs. no DSM-IIIR, DIS Migraine was predictive of 1st inci- Andreski, migraine dent of major depression and panic 1995 disorders Breslau et migraine vs. no DSM-IIIR, DIS Relative risk for major depression al., 1994 migraine associated with prior migraine was nearly the same as the relative risk for migraine associated with prior major depression K. Merikan- migraine vs. DSM III, DSM- Recurrent brief depression, panic gas, 1993b; TTH vs. no IIIR, SPIKE disorder, phobia, generalized anx- K. R. Merikan- headache disor- iety more frequent in migraineurs gas, 1994 der than in all comparison groups Lipton et al., migraine vs. no PRIME-MD Current major depression was sig- 2000 migraine niﬁcantly more frequent in migraineurs than in non-migraineurs Swartz, migraine vs. no DSM III, DSM- Association between migraine and 2000 migraine IIIR, DIS life-time major depression, panic disorder and phobia, but not for substance abuse Breslau et migraine vs. se- DSM IV, CIDI Bidirectional association between al., 2000 vere headache vs. migraine and severe depression no headache disorder Breslau & migraine vs. se- DSM IV, CIDI Bidirectional association found be- Rasmussen, vere headache vs. tween all types of headaches and 2001 no headache dis- panic disorder order 1. Introduction 8

1.2 Pathophysiology of migraine

Next, I will elaborate on the mechanisms assumed to be involved in the genesis and development of a migraine attack. Subsequently, I will focus on the neural network common to migraine and depression.

1.2.1 Mechanisms involved in migraine attacks Current understanding of migraine pathophysiology is still limited. Both, the primary cause of migraine and the mechanisms of pain generation in migraine are incompletely understood. Current views hypothesize that a migraine attack is initiated by a dysregulation in cortical excitability due to various modulatory factors and triggers (genes, gender/hormones, ionic/metabolic, environment). This cortical abnormality may, in turn, be at the origin of spreading cortical waves assumed to be responsible for aura, and contribute to the activation of speciﬁc areas in the brain that promote nociceptive activation of meningeal sensory neurons responsible for migraine pain (Benoit, 2009; Charles & Brennan, 2010). Next, I will describe individually the mechanisms assumed to be involved in the sequence of events leading to migraine attacks (Fig. 1.2).

Fig. 1.2: Integrative model for the sequence of events leading to migraine (adapted from Charles & Brennan, 2010 with added information from Alstadhaug, 2009; Burstein & Jakubowski, 2009). This adaptation of the Charles and Brennan model represents migraine as a feedback loop, where migraine symptoms of sensory sensitivity and mood disturbances project back on the action of extrinsic triggers and modulating factors. The model attempts to encompass the heterogeneous clinical presentation of migraine, by proposing that transitions in the sequence may have discrete thresholds, and connections to distinct molecular, cellular, and neurochemical pathways.

Subject

Mechanisms involved

Extrinsic triggers

Modulating factors

Intrinsic triggers

Sensory Sensitivity (e.g. photophobia)

Brainstem activation:

Trigeminovascular system Dorsolateral pons Brain-Blood-Barrier Nociceptive Head Pain PAG permeability activation Locus coeruleus Environmental (e.g. sensory stimuli) Dysregulation of Cortical waves Hypothalamic & Limbic cortical excitability (e.g. astrocyte waves) activation Nutricional (e.g. irregular food intake) Autonomic Genes Aura Psychological disturbances Ionic/Metabolic (e.g. distress) (e.g. nausea)

Hormonal Mood (e.g. ovulation) disturbances (e.g. anxiety) 1. Introduction 9

Genetic component Genetic load may determine an inherent migraine threshold that is modulated by ex- ternal and internal factors - migraine triggers. The high variability of the migraine phenotype suggests the existence of various genetic components. Genetic studies point to variations in the function of neurons, astrocytes, and vascular cells in different forms of migraine (Haan et al., 2005). Although several susceptibility loci have been found in specific chromosomes, causative genes have not yet been identified, except for the case of familial hemiplegic migraine (FHM), a rare subtype of migraine with aura (Pietrobon & Striessnig, 2003). Namely, mutations in genes involved in FHM ,i.e. FHM1, FM2 and FM3, have been proved to lead to dysregulation of cortical excitability via increased glutamatergic neurotransmission (Moskowitz, Bolay, & Dalkara, 2004). Findings from monozygotic twin studies indicate that migraine with aura (MA) has a stronger genetic influence than migraine-without-aura (MO).

Dysregulation of cortical excitability Substantial clinical data, such as occurrence of symptoms associated with changes of cortical function (e.g. visual aura), and imaging evidence from fMRI and PET studies, indicating dramatic changes of blood ﬂow and metabolic activity in the cortex, support the prominent role of abnormal central neural excitability in migraine pathology. It has been proposed that this cortical dysregulation has an heterogeneous quality that cannot be deﬁned as a simple increase or decrease of excitability, which may contribute to the heterogenous clinical presentation of migraine pathology (Charles & Brennan, 2010). Dysregulation of cortical excitability in migraine is assumed to be the mechanism responsible for triggering propagated cortical waves (Salomone, Caraci, & Capasso, 2009).

Cortical waves Current views suggest that the cortical mechanisms of migraine include a complex interplay between neurons, glia cells, and vascular cells. It has long been suggested that migraine implies a spreading event similar to the Cortical Spread Depression (CSD) observed in animal models. Cortical Spread De- pression is defined as a suppression of electrical activity that spreads slowly across large areas of the cortex (Leao & Morison, 1945 cited in Charles & Brennan, 2010). This propagated wave consists of profound cortical activation followed by sustained inhibition of activity (Aurora & Nagesh, 2010). Results from different methodological studies such as, model of visual stress-induced migraine, spontaneous attacks studies as well as imaging and neurophysiological studies, support the notion that spreading neuronal inhibition affecting the occipital lobe or posterior visual pathway, is in the basis of visual aura in migraine (Benoit, 2009). Recent studies point to the key role of glia cells in cortical activity associated with migraine, based on the discovery that a mutation in an Na+/K+ ATPase primarily ex- 1. Introduction 10 pressed in astrocytes is responsible for familial hemiplegic migraine type 2 (Haan et al., 2005). Although traditionally the astrocytes role in the brain is considered to be mainly of passive and supportive nature, it has recently been proved that astrocytes are capable of extensive intercellular signaling that can modulate both neuronal and vascular activity (Fields, 2009). Astrocytes express a variety of neurotransmitter receptors that allow them to respond to neuronal activity. Conversely, they are capable of active release of transmitters, including glutamate and adenosine thriphosphate (ATP), that can modulate neuronal function (Charles & Brennan, 2010). Astrocytes intercellular communication is primarily achieved by increases in intracellular calcium concentra- tion that are propagated from cell to cell by means of ATP release into the extracellular space and activation of purinergic receptors on adjacent cells. This generates a wave- like pattern triggered by chemical, electrical, or mechanical stimuli, that spreads with temporal and spatial characteristics that are very similar to CSD. Astrocytes are also in close contact with vascular cells, and their signaling mechanisms modulate vascular tone directly through the release of eicosanoids2,K+, and ATP, resulting in either vasoconstriction or vasodilatation (Charles & Brennan, 2010). It is worth mentioning, that due to their close spatial relationship with both neurons and vascular cells, astrocytes are ideal candidates to be responsible for the modulation of widely propagated changes in both vascular activity and neuronal activity observed in migraine (Charles & Brennan, 2010). By themselves, vascular cells are capable of active intercellular signaling, contributing to propagating changes in cortical activity. Early theories regarding the etiology of pain in migraine were grounded on the work of Harold Wolff, who suggested that migraine pain could result from cerebrovascular changes, as most pain structures in the head are vascular-based (Wolf, 1963 cited in Benoit, 2009). CSD may help to understand the complex vascular phenomena in migraine. In mice, spreading depression has an association with a multiphasic vascular response. Which consists in an initial dilatation of cortical surface vessels, followed by a profound constriction of the vessels, followed in turn, by a subsequent vasodilation before the return to normal vascular caliber (Benoit, 2009). Recent functional imaging studies support the existence of significant vasoconstriction in migraine, based on the consistent observation of hypop- erfusion associated with migraine aura, and migraine without aura (Woods, Iacoboni, & John, 1994; Cao, 1999; Géraud, Denuelle, Fabre, Payoux, & Chollet, 2005 cited in Benoit, 2009). A variety of mechanisms are potentially involved in vasoconstriction in migraine: (i) vasoconstriction is associated with the release of messenger an pep- tides that are assumed to mediate migraine pain, e.g. calcitonin gene-related peptide (CGRP) and nitric oxide; (ii) vasoconstriction associated with CSD and/or astrocyte calcium waves could give rise to an uncoupling between blood flow and metabolic activity that in turn could promote a decrease in extracellular pH, triggering a nociceptive response (Benoit, 2009).

2 Eicosanoids are signaling molecules that exert complex control over many bodily systems, mainly in inﬂammation or immunity, and as messengers in the central nervous system. 1. Introduction 11

As described above, cortical waves are associated with the release of a wide variety of neurotransmitters, neuromodulators, and changes in the ionic composition of the extracellular space that can activate nociceptive signaling pathways. Moreover, cortical waves are proposed to inﬂuence changes of permeability in the blood-brain-barrier (Charles & Brennan, 2010).

Brain-Blood-Barrier The blood-brain-barrier (BBB) is a structural and functional barrier that allows only highly selective entry of glucose, aminoacids, and other speciﬁc molecules from the vasculature into the brain via a variety of membrane pumps and transporters. The BBB is comprised of both vascular and astrocytic components, hence propagated cortical activity in migraine could result in changes of BBB permeability. Evidence supporting an increased permeability of the BBB in migraine comes from pharmacological studies, that consistently showed temporal differences in the efﬁcacy of abortive medications based on the phase of the migraine attack in the moment of their delivery (Goadsby, 2009; Charles & Brennan, 2010).

Nociceptive Activation In order to more accurately describe nociception in migraine, I will give an overview of the pain-sensitive structures of the head and its basic pain pathways, before describing the nociceptive pathways in the trigeminovascular system, that are highly involved in migraine. The majority of tissues in the head, face, and neck are pain-sensitive. These comprise the dura, arteries (both intracranial and extracranial), veins, cranial nerves, cervical roots, the periosteum, skin, muscles, sinuses, eyes, ears, teeth, and gums. However, brain parenchyma, pia, and the ventricular lining are not pain-sensitive. Pain sensation from the face and front of the head is transmitted through the trigeminal nerve (V1, V2, V3). The ophthalmic division (V1) of the trigeminal nerve also innervates the pain- sensitive supratentorial intracranial contents. The posterior scalp and posterior and inferior intracranial structures are served by the cervical roots (C2) (Benoit, 2009). Noticeably, the pain tract does not have a speciﬁc cortical mapping, meaning that, neck, face, and head pain is often poorly localized as the ﬁbers of both the cervical and trigeminal systems send mixed signals (Goadsby, 2009; Benoit, 2009). Structures involved in the transmission of trigeminovascular nociception input, and modulation of that input, may be segmented in four categories: (i) target innervation in cranial vessels and dura mater; (ii) a three-neuron pathway comprising the trigeminal complex; (iii) modulatory structures in the midbrain and hypothalamus; and (iv) cortical areas (Table 1.3) (Goadsby, 2009). Several bipolar afferents from dural-vascular structures, corresponding to the target innervations, innervated predominantly by branches of the ophthalmic division of the trigeminal nerve, whose cell bodies are located in the trigeminal ganglion – 1. Introduction 12

Tab. 1.3: Neuroanatomical structures involved in the processing of vascular head pain. Table from Goadsby, 2009

Structure Comments

Target innervation: Ophthalmic branch of – • Cranial vessels trigeminal nerve • Dura mater 1st Trigeminal ganglion Middle cranial fossa

nd 2 Trigeminal nucleus Trigeminal n, caudalis & C1/C2 (quintothalamic tract) dorsal horns 3rd Thalamus Ventrobasal complex Medial nucleus of posterior group Intralaminal complex Modulatory Midbrain Periaqueductal gray matter Hypothalamus Orexinergic mechanisms Final Cortex • Insulae • Frontal cortex • Anterior cingulate cortex • Basal ganglia

first-order-neurons, project on second-order-neurons in the trigeminocervical complex. The trigeminocervical complex extends from trigeminal nucleus caudalis to the caudal portion of the dorsal horn of the C2 spinal cord. Input from trigeminal and cervical neurons converges in the upper cervical dorsal root ganglia into the trigeminovascular complex. Trigeminovascular complex neurons, in turn, project to the thalamus – third-order-neurons and subsequently to cortical areas, including, the somatosensory cortex in the parietal lobe. Nociception modulation is assumed to occur by descending influences onto the trigeminocervical complex, such as those from hypothalamus, midbrain periaqueductal gray (PAG), pontine locus coeruleus (LC), and nucleus raphe magnus (RVM). These influences are assumed to be transmitted via direct and indirect projections. In addition, nociceptive modulation can involve thalamus activation, via LC, PAG, and hypothalamic projections to the thalamus nuclei (Fig. 1.3) (Goadsby, 2009).

Brainstem activation The causal order of activation of cortical and brainstem signaling mechanisms in migraine remains unclear. Functional imaging studies demonstrate activation of the brainstem during migraine attacks. In particular, PET imaging studies revealed increased cerebral metabolism in areas of the brainstem compared to global metabolic 1. Introduction 13 Pathophysiology of Migraine 345

Fig.1. Pathophysiology of migraine. Diagram of some structures involved in the transmission Fig. 1.3: Diagramof trigemin of structuresovascular nociceptive involvedinput inand pathophysiologythe modulation of that ofinput migraine,that forms includingthe trigemi- basis of a model of the pathophysiology of migraine.187 Afferents from dural–vascular struc- novasculartures innervated nociceptivepredominantly inputby transmissionbranches of the first and(ophtha correspondinglmic division) of modulationthe trigem- of that in- inal nerve whose cell bodies are found in the trigeminal ganglion (Vg) project to second put. Vgorder =neuro trigeminalns in the ganglion,trigeminocervical TCCcomplex = trigeminocervical(TCC). The TCC extends complex,from trigeminal DRG= dorsal root ganglia,nucleus PAGcaudali =s midbrainto the caudal periaqueductalportion of the dorsal gray,horn LCof =the pontineC2 spinal locuscord. Input coeruleus, RVM from cervical structures, such as joints or muscle, project through cell bodes in the upper =nucleuscervical raphedorsal root magnusganglia (Figure(DRG) to the fromTCC. Goadsby,TCC neurons 2009)project to ventrobasal thalamus (thalamus) and thence to cortex. Sensory modulation can occur by descending influences onto the TCC that largely respect the midline (dashed line), such as those from hypothalamus, midbrain periaqueductal gray (PAG), pontine locus coeruleus (LC), and nucleus raphe flow (Aurora,magnus Barrodale,(RVM). These Tipton,influences &are Khodavirdi,cartooned as being 2007).direct, but both Thesedirect studiesand indirect also revealed projections are recognized. In addition, sensory modulation can occur from at least LC, PAG, decreased areasand ofhypothalam cerebralic projects metabolismto thalamus innuclei theas medialascending frontalsystems andagain parietalthat largely areas, as well as in the somatosensoryrespect the midline. cortex. This might indicate that migraine entails an abnormality in the inhibitory tone of the higher cortical centers, causing an increased activity in the pain pathways. Though PET is limited in its resolution, authors hypothesized that the activation was in the regions of dorsolateral pons, dorsal raphe nuclei, locus coeruleus and periaqueductal gray (PAG), a structure largely involved in the modulation of ascending nociceptive pathways, with extensive networks from thalamus, hypothalamus, and autonomic nervous system, which could explain the autonomic features of migraine clinical symptoms (Aurora & Nagesh, 2010). Additionally, a specific PET study focusing PAG activation during migraine attacks showed an hyper- activation of this brain area (Bahra, Matharu, Buchel, Frackowiak, & Goadsby, 2001 cited in Aurora & Nagesh, 2010). As seen above by its role in nociception, the trigeminovascular system, integrated with the brainstem, is a central component of migraine pathophysiology. Evidence 1. Introduction 14 supporting the involvement of the trigeminovascular system is grounded in the observation that calcitonin gene-related peptide (CGRP), a neuropeptide known to be involved in cerebral vasoregulation, is increased in jugular venous blood during a migraine attack. It is considered that the release of neuropeptides, i.e. calcitonin gene- related peptide, neurokinin A, and substance P, gives rise to a sterile inflammatory (Raskin, Hosobuchi, & Lamb, 1987 cited in Charles & Brennan, 2010), but the mechanism of pain generation is not yet clear. Moreover, the relevance of the trigeminovascular system in migraine is supported by the existence of binding sites for the serotonin (5-HT) 1B/1D agonists in central terminals of primary afferents within the human brainstem (Goadsby, Lipton, & Ferrari, 2002).

Hypothalamus and limbic activation The hypothalamus presents widespread connections with several areas of the central nervous system. The role of the hypothalamus in the modulation of the autonomic nervous system (ANS) and nociceptive control renders this area a potential intervenient in migraine. A recent PET study focusing spontaneous migraine attacks showed activation of the hypothalamus during the early stages of the migraine crisis, supporting the hypothesis that migraine attacks are caused by a temporary hypothalamic dysfunction (Denuelle, Fabre, Payoux, Chollet, & Geraud, 2007). An alternative explanation of the observed involvement of the hypothalamus in early stages of migraine attacks is that the activation of the hypothalamus constitutes a normal response to distress, given that the hypothalamus plays a pivotal role in psychosomatic responses to stress. Conse- quently, this structure may not be directly related to the genesis of the attack. Conver- sly, a strong argument for the hypothalamic involvement is grounded in sexual dimor- phism. Several hypothalamic sexually dimorphic structures have been found, which could be related to the higher prevalence of migraine in females. Given the central role of the hypothalamus in the control of the circadian patterns, another argument supporting the potential involvement of this structure in migraine is the increased frequency of migraine crisis in association with sleep disorders(Alstadhaug, 2009). Further studies are necessary to correctly identify the role of the hypothalamus in migraine pathogenesis. The amygdala, namely the bed nucleus of stria terminalis, has connections to the paraventricular nucleus (PVN) the neuronal nucleus in the hypothalamus, that may be involved in the mediation of long-lasting behavioral responses during periods of continued distress. Given that these responses persist after the termination of stress, this amygdala-hypothalamic connection has been proposed to explain why migraine may onset both, during and after distress episodes (Alstadhaug, 2009). 1. Introduction 15

Sensory sensitivity Sensitization, i.e. phenomenon where the exposure to an algesic stimulus lowers the threshold for the response of a neuron to the same stimulus, or different stimuli, is often a feature of migraine attacks. Two kinds of sensitization are present in migraine, peripheral sensitization of the primary afferent neuron; and central sensitization of higher-order neurons within the spinal cord and brain. Peripheral sensitization is defined by increased excitability of primary afferents in response to triggers, e.g. mechanical stimulation. Consequently, second-order neurons continuously receive high-frequency impulses from primary afferents innervating the meninges. Inflammatory mediators stimulate the dural receptive fields that directly excite trigeminovascular axons, contributing to their sensitivity. After chemical stimulation, activation of primary afferents is enhanced in the presence of mechanical stimuli that are normally innocuous, resulting in sensitization. It is suggested that sensitization contributes to the hypersensitivity of migraine patients to small changes in intracranial pressure, such as vascular pulsations, resulting in the throbbing quality of a migraine headache (Burstein, 2001; Moskowitz & Buzzi, 2010). Second-order neurons become sensitized after increased inputs from primary afferents. These central neurons within the trigeminovascular complex receive inputs from intracranial meningeal structures as well as from cutaneous structures. It is proposed that central sensitization is responsible for cutaneous allodynia in migraine (Burstein, 2001; Moskowitz & Buzzi, 2010). The existence of central sensitization in migraine was consistently showed in a study in which the pain thresholds of migraineurs to mechanical and thermal stimuli were measured during and between attacks. Findings from this study revealed allodynia in periorbital and forearm skin during the acute migraine attack (Burstein, Yarnitsky, Goor-Aryeh, Ransil, & Bajwa, 2000 cited in Moskowitz & Buzzi, 2010).

Autonomic involvement The autonomic involvement in migraine is clearly expressed by autonomic symptoms present in migraine episodic attacks. However, it is still unsettled if the autonomic nervous system (ANS) plays a role as a causal mechanism in migraine attacks. Autonomic related clinical symptoms of migraine, such as, nausea, vomiting, di- arrhea, bloating, runny nose, and piloerection, are widely accepted to be related with autonomic nervous system dysregulation during migraine. The etiology of autonomic symptoms in migraine is unclear, and distinct pathways are presumably involved, such as the vagus nerve. Specific symptoms of lacrimation and rhinorrhea could be explained by a parasympathetic dysregulation, as the superior salivatory nucleus , situ- ated in the brainstem and integrated with the trigeminovascular system, is activated in migraine (Benoit, 2009). The possible role of the autonomic nervous system in the pathogenesis of migraine has been the focus of several studies using different methodologies, i.e. cardiovascu- 1. Introduction 16 lar reflexes, pharmacological tests, pupillometric and biochemical methods. However the resulting findings have been difficult to explain, since some studies point to hypo- function of the sympathetic nervous system (SNS) as a cardinal feature of migraine while others suggest SNS hyperfunction (see review in Cortelli, Cevoli, Bonavina, Magnifico, & Pierangeli, 2001). In attempt to further the understanding of sympathetic dysfunction in migraine, Peroutka (2004) reviewed the clinical similarity between migraine and other chronic sympathetic nervous system disorders, such as pure autonomic failure and multiple system atrophy. This comparative study draws from pharmacological, clinical and physiological findings that point to a dysfunction of the sympathetic nervous system in migraineurs, presented both in the headache-free period as well as during migraine attacks. Based on noradrenaline-related functional tests, Peroutka proposes that this dysfunction is associated with an imbalance of sympathetic co-transmitters. Particularly, Peroutka suggests that a migraine attack is associated with a relative depletion of sympathetic norepinephrine stores in simultaneous with an increase in the release of other sympathetic co-transmitters such as dopamine, prostaglandins, adenosine triphosphate, and adenosine.

1.2.2 Common neural pathways in migraine and depression In this section I will summarize Burstein and Jakubowski’s (2009) theory of the neural substrate of depression during migraine. Burstein and Jakubowski propose a bidirectional model. On one hand, different migraine triggers related to coping with distress activate multiple brain areas, i.e. hypothalamic, limbic and cortical areas, that con- tain neuronal projections to the trigeminovascular system responsible for prompting the cascade of events that culminate on migraine pain episodes. On the other hand, trigeminovascular projections to speciﬁc areas in the midbrain, hypothalamus, amygdala and basal forebrain are assumed to produce migraine symptoms such as irritability, loss of appetite, fatigue, depression and the pursuit for solitude.

Activation of the trigeminovascular pathway by the limbic system and hypothalamus Burstein and Jakubowski propose that migraine attacks are originating in brain areas that are not directly involved in nociception, but are connected to activate the trigeminovascular pathway, triggering a cascade of events that culminates in migraine pain. A central element of this hypothesis is that migraine triggers related to emotion responses, activate or originate in a number of brain areas whose projections converge on the superior salivatory nucleus (SSN). SSN-projecting neurons located in speciﬁc brain areas ,i.e. bed nucleus of stria terminalis (BNST), paraventricular hypothalamic nucleus (PVN), periaqueductal gray (PAG) are theoretical candidates for mediating the onset of a migraine in relation to emotion triggers (Fig. 1.4). Bed nucleus of stria terminalis (BNST) neurons play a role in the regulation of the hypothalamic-pituitary-adrenal axis, involved in long lasting behavioral responses 1. Introduction 17 S28 Neurol Sci (2009) 30 (Suppl 1):S27–S31

Fig. 1 A proposed parasympathetic pathway for the activation of triggers. b Examples of SSN afferents proposed to be involved in meningealFig. 1.4:nociceptors.HypothesizedPreganglionic parasympatheticparasympathetic neurons pathwayin the migraine leadingtriggering to theby activationolfactory stimuli of meningeal(Pir), food and noci-sleep superior salivatory nucleus (SSN) can trigger intracranial vasodilation deprivation (LH), stress or post stress (PVN, BNST, PAG). BNST and the release ofceptors.nitric oxide SSNin the meninges = superiorthrough salivatorypostganglionic nucleus,bed nucleus SPGstria = sphenopalatineterminalis, LH lateral hypothalamus, ganglion,PAG BNSTperiaq =u- parasympatheticbedneurons nucleusin the sphenopalatine stria terminalis,ganglion (SPG). LHa =The lateraleductal hypothalamus,gray, Pir piriform cortex, PAGPVN = periaqueductalparaventricular hypothalamic gray, SSN receives input from over 50 limbic and hypothalamic brain areas nucleus (red dots) whosePiractivity = piriformmay be influenced cortex,by PVNcommon =migraine paraventricular hypothalamic nucleus. (A) The SSN receives input from over 50 limbic and hypothalamic brain areas (red dots). Activ- to activate theitytrigemi fromnovascula these areasr pathway. is assumedThe trigemino- to be influencedpotassium ions, by commonand action psychologicalof hydrogen ions migrainethrough the vascular pathway consists of first-order nociceptors in the vallinoid receptor (Caterina et al. 1997) or the acid-sensi- trigeminal gangltriggers.ion that (B)inne Representationrvate the meninges; ofsecond- SSN afferentstive ion hypothesizedchannel receptor to(Wa beldmann involvedet al. in1997). migraineConse- order trigeminothalatriggeringmic tract byneuro olfactoryns that receive stimulisensory (Pir), foodquently, andthe sleepactiva deprivationted meningeal (LH),nocicept stressors orrel postease inputs from thestressmeninges, (PVN,periorb BNST,ital skin PAG)and (Figureneck mus- fromcalci Bursteintonin-gene-related & Jakubowski,peptide [ 2009)9] from their peripheral cles; third-order thalamocortical neurons that process branches, resulting in neurogenic inflammation in the dura incoming pain signals from the trigeminal nerve, including [10]. the meninges; and cortical neurons located in the first In contrast to the ongoing effort, to understand how aura somatoduringsenso continuedry cortex. stress, which persist after thetrigge resolutionrs activity in ofmen theingeal stressingnociceptors events., little attenti It ison hypothesized that BNST neurons are involvedwas ingiven migraineto the mechanism triggereds by bywhich distressbrain afterareas the termination of distressful events. involved in regulation of stress could activate meningeal Activation of the trigeminovascular pathway nociceptors and trigger the headache. Is there a common by theParaventricularlimbic system and hypothal hypothalamicamus nucleus (PVN)pathway neuronsthat activates thatmen projectingeal noci toceptors sympatheticfor a variety and parasympathetic preganglionic neuronsof inmigra theine brainstemtriggers? We are andpropos spinaling that cordsuch a arepath in-way Thevolvedobservation in thethat autonomicvisual aura responses precedes the toonset stress,of potentiallyinvolves pre- and includingpostganglionic localizedparasympat cerebrovas-hetic neuron headache by several minutes promoted extensive research in the superior salivatory nucleus (SSN) and sphenopala- oncularthe neura vasodilationl substrate inby thewhich initialcortical phasespreadi ofng the migrainetine ganglion attack.(SPG), respectively. According to our depressiPeriaqueductalon can result in activa graytion of(PAG)meningeal neuronsnocicep- arehypot assumedhesis, migra toine betrigge involvedrs either inactiva passivete or orig emo-inate in tors. Evidence suggests that in the wake of cortical a number of brain areas whose projections converge on the spreaditionng, copingdepression withthe unavoidableblood brain barrie stressorsr becomes that influenceSSN. The SSN, thein increaseturn, stim ofulates migraine the release episodeof acetyl morefrequency permeable associated[6, 7], allowing withpotassium longand periodshydrogen of socialcholine distress., vasopressin intestinal peptide and nitric oxide ions to diffuse from the surface of the cortex to the pia from meningeal terminals of SPG neurons, resulting where they activate C-fiber meningeal nociceptors [8]. This (directly or indirectly) in a cascade of events that include activaActivationtion appea ofrs to theinvolve hypothalamusdirect depolarizati andon limbic by systemthe dilation byof theintracran trigeminovascularial blood vessels, plasma pathwayprotein

123According to Burstein and Jakubowski, pain signals that originate in the trigeminovascular pathway during migraine modify the activity of hypothalamic and limbic structures involved in the integration of sensory, physiological and cognitive signals driving 1. Introduction 18

behavioral, affective and autonomic responses. It is proposed that ascending pathways from neurons located in the ventrolateral area of the upper cervical and medullary dorsal horn (an area containing the majority of second-order trigeminovascular neurons projecting to speciﬁc brain areas, i.e. lateral hypothalamus, paraventricular nucleus of the hypothalamus and forebrain nuclei) is functionally positioned to produce irritability, loss of appetite, sleepiness, fatigue, chill, stress, depression, emotion arousal, decreased motivation, the pursuit for solitude and lethargy during migraine (Fig. 1.5). Neurol Sci (2009) 30 (Suppl 1):S27–S31 S29

Fig. 2 Proposed mechanism for the initiation of symptoms commonly associated with migraine headache by ascending trigeminovascular pathways to the brainstem, hypothalamus and basal ganglia. a Trigeminovascular neurons in the spinal trigeminal nucleus (SpV) project to multiple limbic and hypothalamic brain areas (red dots) whose activity my underlie common migraine symptoms. b Examples of SpV projections proposed to be involved in stress (PVN), decreased motivational state (VP/SI), pursuit of solitude (PAG), sleepiness, irritability and loss of appetite (LH). LH lateral hypothalamus, PAG periaqueductal gray, PVN paraventricular hypothalamic nucleus, VP/SI ventral pallidum/substantia innominata

extravasation, andFig.local 1.5:releaseHypothesizedof inﬂammatory mechanismmolecules forParvoc theellular genesisPVN ofneurons depressionthat project symptomsto sympathetic commonlyand associ- that activate adjacent terminalsatedof withmeninge migraineal nociceptors headacheparasym by ascendingpathetic prega trigeminovascularnglionic neurons in the pathwaysbrainstem to the brain- (Fig. 1). stem, hypothalamus and basaland spinal ganglia.cord promot SpVe =the spinalautonomic trigeminalpart of the nucleus,stress LH = lat- Several lines of evidence eralsuppor hypothalamus,t this parasympathetic PAG =respon periaqueductalse [29, 30], which gray,include PVNs localize = paraventriculard cerebrovascular hypothala- hypothesis: (1) meningeal blood vessels are densely vasodilation in the early phase of the migraine attack [31]. innervated by parasympatheticmicﬁbers nucleus,[11–13]; VP/SI(2) pregan- = ventralVentr pallidum/substantiaolateral PAG neurons invol innominata.ved in passiv (A)e emoti Trigeminovascularonal glionic parasympathetic neuronsneuronsin the SSN locatedincreas ine thetheir spinalcoping trigeminalwith inescapa nucleusble stress (SpV)ors such projectas repeate tod defeat variousin limbic and activity after activation of menhypothalamicingeal nociceptors brain[14]; areas(3) (representedsocial encounters in[32 the, 33 diagram] may mediate by redonset dots)of incr whoseease activity is ongoing activity in meningeal nociceptors appears to migraine frequency associated with a long period of social depend on enhanced activityassumed in the SPG to[15 underlie]; (4) para- commonstress migrainesuch as divorce. symptoms. (B) SpV projections proposed to sympathetic tone is enhancedbedu involvedring migra inine, stressas evi- (PVN), decreased motivational state (VP/SI), pursuit of soli- denced by lacrimation, teary eyes,tudenasal (PAG),congesti sleepiness,on [5]; (5) irritability and loss of appetite (LH) (Figure from Burstein & blockade of the SPG providesJakubowski,partial or complete 2009)relief of Activation of the hypothalamus and limbic system by migraine pain [16–25]. the trigeminovascular pathway The SSN receives extensive input from more than 50 brain areas distributedTrigeminovascularthroughout the forebrain, projectionsdien- The tomost thefrequently lateral hypothalamusreported symptoms mayassociated be responsiblewith for cephalon, midbrain, pons and medulla [26]. SSN-project- migraine are depression, stress, irritability, fatigue, sleepi- ing neurons locatedsymptomsin some o suchf these asbrain lossareas ofare appetite,ness, exagge sleepinessrated emotional and irritabilityresponses, nause duringa and loss migraine.of The theoretically positionedhypothalamusto mediate the hasonset widespreadof a migraine projectionsappetite. To elici tot thethese cerebralsymptoms, cortex,pain sign andals that isorig- assumed to be by means of theirinvolvedinvolvemen int the in emotional regulationresponses of foodinate andin waterthe trigem intake,inovascular sleeppathway and arousal.during migraine (Fig. 1a). The bed nucleus of stria terminalis (BNST), the must reach and alter the activity of hypothalamic and paraventricular hypothalamicTrigeminovascularnucleus (PVN) and projectionsthe PAG limbic to thestru paraventricularctures that integrate nucleussensory, physi of theological hypothalamus are all involved mayin the playcircuitry a centralthat regulates role in‘‘stress migraineand cognitive associationsignals withthat drive distress,behavior sinceal, affective thisand nucleus con- response’’. BNST neurons, which regulate hypothalamic- autonomic responses. Brain areas involved in the execution pituitary-adrenal axis, appear to mediate long-lasting of such responses include the parabrachial complex, PAG, behavioral responses during sustained stress, which persist hypothalamus, amygdala, septum, nucleus accumbens, bed long after the termination of stress [27, 28]; such neurons nucleus of the stria terminalis and basal ganglia [34–46]. may be involved in stress-induced migraine and also Many of these brain areas receive direct inputs from lam- in migraine triggered after the termination of stress. inae I–II and V neurons located in the ventrolateral area of

123 1. Introduction 19 tains neurons expressing corticotrophin releasing hormone and oxytocin involved in regulating stress responses. Trigeminovascular projections to forebrain nuclei, i.e. ventral pallidum and substantia innominata, may mediate emotion arousal and decreased motivation during migraine. These areas contribute to the modulation of endocrine, autonomic and so- matomotor functions in relation with emotion and motivational states. Trigeminovascular projections to PAG may contribute to behaviors of pursuit of solitude during migraine, as it as been showed that ventrolateral PAG neurons are involved in mediating responses to deep, inescapable pain. Burstein and Jakubowski propose a network of bidirectional signaling in migraine where the trigeminovascular pathway is mutually connected with limbic and hypothalamic systems. This bidirectional mechanism may position the emotion dimension in migraine as one of the main factors contributing for the duration of the migraine attacks, which can last for many hours and even days.

1.3 Migraine treatment approaches

Bellow, I describe the main pharmacological and non-pharmacological approaches to migraine treatment. Subsequently I will elaborate on the topic of emotion modulation of migraine pain, highlighting strategies of emotion reappraisal of subjective pain experience and the action of descending modulatory circuits of nociception which integrate prominent emotion related structures.

1.3.1 Available approaches Migraine treatment can be divided into non-pharmacological and pharmacological approaches. Often patients follow treatment that include pharmacological as well as non-pharmacological measures. The most common non-pharmacological approach is patient education. This method aims at promoting life-style changes that allow migraineurs to better identify and avoid migraine triggers, these include, development of regulatory habits such as regular sleep, regular meals, exercise, avoidance of distress, relaxation techniques, and avoidance of diet related triggers. However, the sensitivity of the brain to such triggers at any given time is uncertain and for some patients this kind of approach reveals to be frustrating, since it will lead to different outcomes on different days (Goadsby et al., 2002; Moskowitz & Buzzi, 2010). This indicates that behavioral strategies need to be deeper than the supply of information and medical advices. Other existent methods such as those grounded in cognitive behavioral therapy and biobehavioral training, such as biofeedback (Vasudeva, Claggett, Tietjen, & Mc- Grady, 2003; Andrasik & Rime, 2007; Nestoriuc, Martin, Rief, & Andrasik, 2008), relaxation-training (Stetter & Kupper, 2002; Lawler & Cameron, 2006; John, Sharma, Sharma, & Kankane, 2007), and stress management (Kraft, Lumley, D’Souza, & Doo- 1. Introduction 20 ley, 2008) represent promising alternatives to approach migraine treatment and management without recurring to pharmacological treatment (Buse & Andrasik, 2009). These approaches have demonstrated favorable results when learned and practiced correctly, and may be used on their own or in combination with pharmacological in- terventions. Pharmacological approaches can be divided into prophylactic, i.e. preventive drugs that are taken daily, whether or not the headache is present, to reduce the frequency and severity of attacks, and acute, i.e drugs that are taken to treat attacks when they arise. Pharmacological prophylactic treatment generally consists in: amitriptyline with a pharmacological action that inhibits sodium channels; flunarizine that blocks calcium channels; beta-blockers i.e. propranolol and timolol, that block noradrenaline receptors; or anticonvulsants ,i.e. divalproex and topiramate which complex action involves modulating ion channels as well as gamma-aminobutyric acid, glutamate, and kainate receptors. The choice of the appropriate drug has to take into consideration contra-indications, comorbid conditions and market availability. Most preventive medications are thought to act, partially, by normalizing neuronal activity and increasing the neuronal discharge threshold. Namely, the effectiveness of calcium channel blockers or anticonvulsants is based in the suppression of central hyper excitability via the promotion of changes in voltage-gated channels. Conversely, for serotonin reuptake inhibitors and beta-blockers, the central mechanism of action is assumed to be the modulation of neurotransmitter release (Moskowitz & Buzzi, 2010). Pharmacological treatments for the acute phase of migraine can be divided into nonspecific and migraine-specific treatments. Nonspecific treatments, also used to treat other pain disorders include analgesics such as aspirin, acetaminophen, non- steroidal anti-inflammatory drugs, opiates, and combination analgesics. Specific treatments, used with efficacy in the treatment of neurovascular headaches, include ergo- tamine, dihydroergotamine, and the triptans. The pharmacological mechanism of these drugs is based on a vasoconstriction action promoted directly or indirectly (Goadsby et al., 2002). Some migraine-specific drugs, including triptans, are effective due to their serotonin agonist role at the 5HT1 receptors. Triptans are potent agonists of the 5- HT1B and 5-HT1D receptors, and some bind efficiently to 5-HT1F receptors. This allows them to act on specific 5-HT receptor binding sites expressed by the trigeminovascular system, an area highly involved in migraine nociceptive mechanisms. Moreover, some available triptans are capable of crossing the blood-brain barrier at therapeutic doses, augmenting their success in the treatment of migraine attacks (Moskowitz & Buzzi, 2010). The variety of triptans available (almotriptan, eletriptan, frovatriptan, naratriptan, rizatriptan, sumatriptan, and zolmitriptan) contributes to a more individually tailored treatment (Moskowitz & Buzzi, 2010). Triptans may be indicated for the treatment of migraineurs suffering from depression comorbidity, because of their action on the serotonergic system, a system that plays a prominent role in the pathogenesis of mood disorders. Investigation of the role of serotonin in headache is an area of intense interest (Benoit, 2009). 1. Introduction 21

1.3.2 Emotion modulaion In the following, I will summarize the current cognitive neuroscience approach to emotion, focusing on the biphasic approach, on which the experimental paradigm of this thesis is based. This will be followed by a overview of the methods and limitations of inducing and measuring emotion in the context of scientiﬁc investigations. After- ward, I will elaborate on the mechanisms subserving cognitive modulation of pain, focusing on emotion mechanisms. Finally, I will address current research on emotion modulation of pain and nociception in headache disorders.

Emotion Emotion research finds one of its main challenges in the actual definition of what an emotion is. Some say that there are almost as many definitions as there are researchers in the field (see Panksepp, 1982 for a representative list, and Prinz, 2006 for a critical review of emotion theories). Emotion involves complex patterns of cognitive, affective, behavioral and physiological changes produced automatically by the brain in response to intrinsic and extrinsic stimuli. Functionally, emotion is assumed to constitute action dispositions that may or may not pass through conscious processing, mobilizing an individual for a specific behavior. In the context of this thesis I will use a working definition of emotion, that posits emotion as a mode of reaction of brains that evolved to respond to certain classes of stimuli with certain repertoires of action directed at achieving general goals ultimately increasing the inclusive fit- ness. This stimuli-repertoire-goal interaction is extended in an idiosyncratic fashion as individuals learn to respond in adaptive ways to the large variety of circumstances in their environment (Damasio, 2001). Most authors are making a clear distinction between feelings and emotion (Damasio, 2001; Damásio, 2003). Feelings are direct consequences of emotion; they comprise a mental representation of the cognitive, affective, behavioral and physiological responses that characterize emotion. Whereas emotion supply an immediate response to changing environmental demands faced by an organism, feelings provide a mental awareness of emotion (as a particular kind of higher-level computational product of the factors concurring to produce emotion responses). Moreover, feelings have a strong functional importance directly related to memory and learning.

Cognitive neuroscience approach to emotion From a neurobiological perspective, emotion is a product of the integration of central nervous system (CNS) and autonomic nervous system (ANS) activation. Current research in affective neuroscience seeks to understand the central nervous system networks that underlie emotion. Inves- tigations focusing on the CNS correlates of emotion have identified the involvement of multiple cortical (e.g. frontal, temporal, and parietal) and subcortical areas (e.g. basal ganglia, thalamus, amygdala, and hippocampus) in a wide variety of positive and negative emotions (Lane & Nadel, 2002). The activation of these circuits have widespread 1. Introduction 22 effects on other cortical functions leading to cognitive adjustments, such as changes in attention as a response to an emotion eliciting stimulus. Psychophysiological research strives to identify the autonomic nervous system patterns that correlate with emotion states. Emotion responses targeting the ANS result in the emergence of bodily emotion states, that encompass dynamic and homeostatic internal adjustments, and contribute toward specific behaviors, e.g. freezing or fight-or- flight, and facial expressions. One main trend pursued in the investigation of the ANS role is the search for emotion specific differences in physiological activation patterns during distinct emotion states. However, this approach has shown that there is a higher amount of similarities in physiological activation patterns, than specific differences between emotion states (for review Cacioppo, John; Tassinary, 2007). Recently proposed emotion models, capturing the attention of the neuroscientific community, include a dynamic integration of CNS and ANS, e.g. neuron-visceral model (Hagemann, 2003) and embodied models (Craig, 2002; Critchley, 2005). The fundamental principles guiding these models draw inspiration from dynamical systems concepts, proposing that several neural networks are flexibly recruited in accordance to situational demands, giving rise to integrated central and autonomic responses to emotion stimuli.

Biphasic approach to emotion In this project I will consider emotion from a biphasic motivational perspective. Subsequently I will briefly describe the guiding concepts of this theory. The biphasic approach assumes that a primary distinction between emotion events is “hedonic”, i.e. good or bad, appetitive or aversive, agreeable or disagreeable, positive or negative, pleasant or unpleasant, hospitable or inhospitable (Cacioppo, John; Tassinary, 2007). Moreover, events differ in the amount to which they arouse or activate responses (Russell, 1980). The above described dimensions, that are commonly called “valence” and “arousal”, are clearly related to the motivational parameters of direction and intensity found in animal behavior theories. Em- pirical studies show that even in simple organisms, stimuli that promote survival elicit behaviors towards the eliciting stimulus, whereas stimuli that threaten the organism prompt withdrawal, and that these behavioral responses occur with different intensi- ties. Due to the link between emotion and motivation, a number of emotion theorists ad- vocate a biphasic approach to emotion. The biphasic theory of emotion posits that human emotion emerge from varying activation in neural circuitry of appetitive/approach and defensive/avoidance. This circuitry, integrating areas from the cortex, sub-cortex, and midbrain, is thought to have been largely defined early in evolutionary history to mediate behaviors basic for the survival of individuals and the propagation of genes to coming generations (Davidson, 2003; P. J. Lang & Davis, 2006). According to the biphasic approach to emotion, discrete states of human emotion, e.g. joy, surprise, fear and sadness, are tactical responses similar to the long observed animal strategies, e.g. freeze, fight. Although in humans, the execution of these behavioral strategies is no 1. Introduction 23 longer completely dictated by the eliciting stimulus itself, basic motivational parameters of direction (towards, away) and intensity can still be considered fundamental to organizing emotion responses and states. In support of this several studies show activation patterns in brain and body in response to emotion cues, co-varying with participants’ reports of affective valence and increasing emotion arousal (P. J. Lang & Davis, 2006). Furthermore, P. J. Lang (1995) proposed the motivational priming hypothesis, which predicts that an organism’s emotion state modulates responses to emotion significant stimuli. Meaning that, responses triggered by aversive stimuli are facilitated in the context of a negative emotion state and inhibited in the context of a positive emotion state. This prediction was strongly verified by several studies focusing startle reflex in animals and humans (P. J. Lang & Davis, 2006).

The scientiﬁc study of emotion has to devise valid methods of inducing and measuring emotion in the laboratory context. Next, I will address the main methods used in the scientiﬁc study of emotion.

Emotion induction The study of emotion in the scientiﬁc realm usually starts by the induction of affective responses. It is necessary a controlled induction methodology in order to accomplished valid measurements of affective responses. Common contexts of inducing affective reactions can be organized into those that primarily target perception, anticipation, imagination, or action. These are not mutually exclusive (e.g. anticipatory tasks that include perceptual input). Perception tasks focus on measuring affective responses to sensory stimuli that is presented in several modalities, i.e. visual (e.g. pictures, ﬁlms), acoustic (e.g. music), tactile (e.g. shock), olfactory, or gustatory. Imagery tasks aim at affective responses in the context of mentally portrayed affective events. Affective induction by anticipation methods, targets reactions assessed while a subject awaits the presentation of an affective stimulus (e.g. threat of shock studies and classical conditioning paradigms). Action induction contexts include situation in which emotion responses can be measured (e.g. giving a speech) (Cacioppo, John; Tassinary, 2007).

Measuring emotion Standardized methodology to measure emotion comprise evaluative self reports, recording of overt actions, and measurement of physiological responses. In much of psychological research, emotion measurement relies mainly on evaluative self reports, including open-ended verbal descriptions (e.g. “I’m afraid”), Likert scale ratings (e.g. ratings of fear on a scale of 1 to 10), and reports of associated responses (e.g. indicating or listing bodily reactions). Overt actions, such as running, jumping, fighting, freezing, etc., used extensively in studies of motivated behavior in animals, are less commonly measured in human studies, with the exception of facial expressions which were studied extensively over the past decades (for review Ekman, 2009). Physiological responses are bodily events that can be assessed using psychophysiological and neurophysiological instrumentation and methods. The for- 1. Introduction 24 mer includes the recording of electrodermal activity (EDA), cardiovascular responses, electrical activity of the heart (electrocardiography – ECG), respiration, and neuroen- docrine responses. The latter – neurophysiological measurements – include recoding changes in electrical activity of the brain (electroencephalography – EEG), cerebral blood flow (functional magnetic resonance imaging – fMRI) and cerebral metabolism (positron emission tomography – PET). If emotion is defined in terms of subjective reports, overt behaviors, and bodily responses, one might wonder which measure, if any, best captures conscious feelings. From a measurement perspective, one solution is to operationally define conscious experience on the basis of evaluative reports: Subjects’ reports (verbal or non-verbal) about their emotion experience can be used to index internal state that is usually meant when one speaks of feelings. Some might argue this is unsatisfactory, due to the depen- dence of personal reports on cultural norms and individual differences in disclosure.

Emotion influence in nociception and pain The International Association for the Study of Pain (IASP) defines pain as “an unpleasant sensory and emotion experience associated with actual or potential tissue damage or described in terms of such damage” (Loeser & Rolf-Detlef, 2008). Nociception is defined as “the neural processes of encoding and processing noxious stimuli” (Loeser & Rolf-Detlef, 2008). Hence, this currently accepted terminology suggests that pain is not exclusively driven by the noxious input, but is a complex function of the human brain that involves emotion processes. In the following, I will describe the mechanisms supporting cognitive modulation of pain, and present studies that support the influence of emotion in nociceptive modulation. Subsequently, I will address recent research focusing on the contribution of positive emotion induction in pain attenuation.

Emotion aspects in the subjective experience of pain Three main mechanisms are assumed to lead to cognitive modulation of pain: attention, expectation and reappraisal (Wiech, Ploner, & Tracey, 2008). Emotion modulation of pain dynamically integrates attention and expectation mechanisms. It is assumed that cognitive modulation of pain through attention mechanisms occurs by means of a distraction process, where com- petition for attention between a highly salient sensation – pain, and an intentionally directed focus on another information activity, results in the attenuation of the experience of pain (Wiech et al., 2008). Emotion salient stimuli are shown to cause strong attentive attraction (Compton, 2003), suggesting that a distraction based process is also involve in pain modulation through emotion stimulation. Most studies addressing emotion modulation suggest that a valence-by-arousal interaction best characterizes the inﬂuence of emotion on pain (Rhudy, Williams, Mc- Cabe, Nguyen, & Rambo, 2005). Findings from several experimental studies have consistently shown that positive emotion, including experimentally induced positive mood, leads to reduced pain in healthy individuals. Furthermore, the degree of inhibition is determined by the intensity of the emotion state (assessed by arousal), with more 1. Introduction 25 intense positive emotion, e.g. sexual excitation, eliciting the greater inhibition (Rhudy, Williams, Mccabe, Russell, & Maynard, 2008). In contrast, negative emotion with low-to moderate intensity arousal ,e.g. anxiety, lead to enhanced pain, whereas negative emotion with high arousal, e.g. fear, lead to decreased pain (Kenntner-Mabiala & Pauli, 2005). This differences in emotion induced analgesia can be related to the involvement of expectation mechanisms in the modulation of pain. According to the motivational priming hypothesis, an organism’s emotion state will modify responses to signiﬁcant stimuli. Responses triggered by aversive stimuli are facilitated in the context of a negative emotion state and inhibited in the context of a positive emotion state (P. J. Lang, 1995). In this way, negative emotion stimulation that is not strong enough to compete with pain for attention resources may result in pain enhancement, by promoting a negatively driven reappraisal of the nociceptive stimuli.

Emotion mechanisms integrating nociceptive modulation Next I will give a brief overview on the complex phenomenon of nociceptive processing in order to focus on the descending modulation of nociception and how it is linked to emotion mechanisms. In general, nociceptive information from the viscera, skin and other organs is subject to extensive processing by several mechanisms that modulate nociception by in- hibiting or enhancing nociceptive signaling to higher centers, e.g. somatosensory cortex (SSC). In this regard, a network of descending pathways projecting from cerebral structures to the dorsal horn (DH) plays a complex and crucial role. Specific pathways either suppress, descending inhibition (DI), or potentiate the passage of nociceptive messages, descending facilitation (DF). Noticeably, there is no concrete anatomical separation of substrates involved in descending processes of inhibition and facilitation, and the activation of a single supraspinal structure may trigger both DI and DF. Trans- mitters typically contributing to descending inhibition are e.g. cannabinoids, adenosine and nitric oxide. Transmitters playing a role in descending facilitation include e.g. substance P, calcitonin gene related peptide, glutamate, noradrenaline, GABA, en- dorphin, vasopressin and oxytocin (Millan, 2002). Dysregulation of this descending modulatory systems may lead to a decrease in nociception inhibition, an increase in nociceptive facilitation, or both, resulting in an increased nociception transmission. In this case, these supraspinal modulatory systems that in normal conditions help main- tain “pain balance”, might be involved in the development of chronic pain disorders, including headache disorders (Benoit, 2009). Findings coming from animal studies reveal ascending nociceptive and descending modulatory pathways that may contribute to the prominent role of affective and motivational aspects in the modulation of nociception (see review Villemure & Bushnell, 2002. In humans, imaging research reveals a congruent pattern of cerebral activity that occurs during the subjective experience of pain. This activation pattern includes the amygdala, hypothalamus, periacqueductal gray (PAG) and anterior cingulate cortex (ACC) (Rhudy & Meagher, 2001; Roy, Piché, Chen, Peretz, & Rainville, 2009). Fur- ther experimental evidence for emotion related descending modulation of pain comes 1. Introduction 26 from studies showing the effects of electrical stimulation of the PAG and rostroventral medulla (RVM) on nociceptive spinal neurons (Porreca, Ossipov, & Gebhart, 2002). These studies revealed that electrical brainstem stimulation did not induce extra activity in the spinal nociceptive neurons when they were not being stimulated, but it modulated their response magnitude to stimulation of cutaneous and visceral receptive fields in response to nociceptive and innocuous stimuli. Relying on these findings, a circuit involving the anterior cingulate cortex (ACC), hypothalamus, amygdala, periaqueductal gray (PAG), and rostral ventromedial medulla (RVM) is proposed to be highly involved in cognitive and affective modulation of nociception (Rhudy & Meagher, 2001). Noticeably, these brain structures coincide with the model of depression during migraine proposed by Burstein and Jakubowski, supporting the hypothesis of the prominent involvement of emotion mechanisms in the genesis and persistence of migraine pain.

Contributions of positive emotion induction to pain and nociception attenuation in migraine The observed association of emotion and pain mechanisms has inspired research focusing on the contribution of positive emotion induction (achieved by visual stimuli) in pain attenuation, suggesting a route for the development of novel biobehavioral therapeutic approaches to migraine (Williams & Rhudy, 2009c, 2009a; Tommaso et al., 2009). In the following, I will briefly elaborate on the findings from these studies. A study by Williams and Rhudy (2009c) probed the supraspinal modulation of trigeminal nociception and pain in healthy subjects. To assess modulation of trigeminal nociception is highly relevant for migraine research, due to the prominent role of the trigeminalvascular system in migraine pain. Findings from this study showed that emotion induction through affective pictures effectively modulates trigeminal nociception and pain (measured by nociceptive blink reflex and pain ratings respectively). In particular, this study suggests that pleasant stimuli inhibited both trigeminal nociception and pain, whereas unpleasant stimuli enhanced trigeminal nociception and pain (Williams & Rhudy, 2009c). A subsequent study by the same authors Williams and Rhudy (2009a) focused emotion modulation of autonomic responses to painful stimulation through affective pictures. This investigation showed a linear trend in emotion modulation of pain- evoked heart-rate (HR) and electrodermal activity (EDA). Pleasant conditions showed lower autonomic activation than neutral ones, and autonomic activation in neutral conditions was lower than in unpleasant conditions. This observation is consistent with the motivational priming theory integrated in the biphasic motivational theory of emotion. Considering EDA and HR acceleration as responses to defensive activation, the motivational priming theory predicts that defensive responses are facilitated by stimuli that prime the defensive system (viewing unpleasant pictures) and inhibited by stimuli that prime the appetitive system (viewing pleasant pictures). A study by Tommaso Tommaso et al. (2009) addressed the effects of affective 1. Introduction 27 pictures on pain sensitivity (measured by subjective ratings) and cortical responses induced by laser stimuli (measured by laser-evoked-potentials (LEP), and recorded with electroencephalogram (EEG)) in healthy subjects and migraineurs without aura. Find- ings showed that stimulation by images reduces subjective pain independently of the valence and arousal qualities of the picture stimuli. This suggests that the mechanism directing cognitive modulation of pain is fully driven by attention processes indepen- dent of the quality of the affective stimuli. The authors explain that differences in procedure may explain the discrepancy between this finding and results from previous studies that showed a valence-by-arousal interaction in emotion modulation of pain (Rhudy et al., 2005; Kenntner-Mabiala & Pauli, 2005; Rhudy et al., 2008). However, a special inhibition of laser-evoked-potentials was observed in positive emotion conditions, suggesting a possible interference of positive affective induction in the mechanisms that lead to nociceptive modulation. In migraineurs, affective pictures were able to modulate pain perception and laser-evoked-potentials differently from other modes of distraction, supporting emotion induction as a potential approach towards cognitive modulation of migraine pain (Tommaso et al., 2009). The preliminary evidence described above suggests that emotion manipulation may be a promising approach to migraine treatment. On one hand, positive emotion induction acts over the cognitive elaboration of subjective pain, by promoting the reappraisal of pain in less aversive contexts and offering a valid distraction stimulation. On the other hand, emotion manipulation seems to be able to influence nociceptive modulation via descending mechanisms, e.g. PAG-RVM pathways, resulting in a measurable inhibition of nociception. Migraine pathology reveals a specific association with emotion mechanisms that is expressed in several dimensions, i.e. comorbidity between migraine and psychiatric disorders of depression and anxiety (Torelli & D’Amico, 2004; Radat & Swend- sen, 2005; Frediani & Villani, 2007), migraine association with specific personality traits (K. Merikangas, 1993b; Fan et al., 1999; Bag et al., 2005; Tan et al., 2007; Sánchez-Román et al., 2007; Abbate-Daga et al., 2007), the prominent role of distress as a trigger to migraine attacks (Schoonman et al., 2007; Levy et al., 2009; Wober & Wober-Bingol, 2010), and actual mood abnormalities experienced during migraine attacks (Benoit, 2009; Burstein & Jakubowski, 2009). The strong association between migraine and emotion in addition with recent findings on emotion modulation of pain and nociception related to migraine, render migraine pathology an ideal candidate for the development of emotion-based treatment approaches. While the studies mentioned above lay foundation for assessing the viability of novel therapies focusing on emotion modulation of migraine pain and nociception, a detailed understanding of migraineurs emotion profile is still required and may contribute to a better design of emotion-based therapeutic approaches. 1. Introduction 28 2. METHODS

In the following, I will describe the methods of the empirical study conducted, which aimed at examining speciﬁc patterns of affective responses in migraineurs with aura. In addition, I will elaborate on the contributions of this thesis to the development of the Affective Multimodal Data Base system (AMDB).

2.1 Experimental design

In the following I will outline the specific goals that were taken into account in the design of the experimental protocol, and describe the temporal structure of the experiments. To assess the affective profile of migraineurs with aura, I designed an experimental protocol composed of a “Single picture phase” and “Transient pictures phase” phase. In the first phase, “Single picture phase”, I presented pictures individually and asked participants to rate how they felt after each picture presentation. The goals of the “Single picture phase” were: • Probe differences in subjective responses to affective pictures between migraineurs with aura and non-migraineurs. With this goal I expected to replicate and extend findings of enhanced emotion impact in migraine for both pleasant and unpleasant pictures, attained by the only previous study dealing with the effects of the International Affective Picture System pictures (P. Lang, Bradley, & Cuthbert, 2008) in migraine patients (Tommaso et al., 2009); • Provide baseline data for the second phase of the experiment. In particular, data from the first phase is used to better estimate the valence-inhibition effect observed in the affective responses to transient stimuli that compose the second phase of the experiment; • Inspect trends in affective responses due to repetitive presentation of affective stimuli. Given the prediction that the sequential presentation of affective stimuli over a period of time may lead to effects of habituation and/or potentiation of affective responses that may differ between migraineurs with aura and non- migraineurs In the second-phase, “Transient picture phase”, I presented transient stimuli composed of a sequence of two pictures. I considered the first picture presented as the context, setting an initial affective state, and the second picture presented as the inhibitor, modulating the affective state induced before. After the presentation of each transient stimuli I asked participants to rate how they felt. The goals of the “Transient 2. Methods 30 pictures phase” were: • Probe differences in the flexibility to adapt affective responses to shifts in emotion stimulation, as measured by the valence-inhibition effect; • Inspect if differences in transition timings influence affective responses. The ratings for both phases were done with two Likert scales, one scale assessing valence, i.e. the degree of pleasantness, the other measuring arousal, i.e. the intensity of emotion experience. Both Likert scales ranged from 1 to 9.

2.1.1 Temporal structure of the phases The ‘‘Single picture phase” was initiated by a baseline period of two minutes, during which participants were asked to relax. This was followed by 30 trials of single picture presentations. Each trial was composed by the same sequence of events, starting with a blackout (4 ± 1 seconds) followed by the picture presentation (6 seconds) and a subsequent blackout (3 seconds). After each trial participants performed the rating task (9 seconds) (Fig. 2.1). In total six pictures of each picture class (high-arousal- pleasant, low-arousal-pleasant, high-arousal-unpleasant, low-arousal-unpleasant, and neutral) were randomly selected and presented. A randomization procedure for the order of presentation was applied, to achieve a balanced amount of the different picture classes between the first half and the second half of the “Single picture phase”. This was included in the design to allow a better assessment of habituation and/or potentiation effects. The “Transient pictures phase” immediately followed the first phase, and was initiated by a new baseline period of two minutes, during which participants were again instructed to relax. After this period 36 trials consisting of transient stimuli were presented. Each trial comprised the same sequence of events. The trials from the second phase of the experiment started with a blackout (4 ± 1 seconds), followed by the presentation of the context picture (6 seconds) and a subsequent blackout interval (1, 3, or 6 seconds), after which the inhibitor picture was presented (6 seconds) followed again by a blackout interval (3 seconds). Each trial was succeeded by a rating task (9 seconds) (Fig. 2.1). Trials were randomized taking in to account two rules (i) all the transient stimuli classes are presented, and (ii) transient stimuli belonging to each class are integrated in equal amounts in the first and second half of the “Transient pictures phase”. The overall duration of the complete experiment was approximately thirty-two minutes, eleven minutes for the first-phase and twenty-one minutes for the second-phase. To avoid pre-exposure effects, visual stimuli used throughout the experiment were only presented once to each participant.

2.2 Stimuli

As pictorial stimuli one-hundred-and-fourteen color slides from the International Af- fective Picture System (IAPS) (P. Lang et al., 2008) were chosen. The choice of the 2. Methods 31

Single Picture Phase

Baseline Black Single Black Subjective Loop Picture Ratings (x30) (120 s.) (3/5 s.) (6 s.) (3 s.) (9 s.)

Transient Pictures Phase

Baseline Black Context Black Inhibitor Black Subjective Loop Picture Picture Ratings (x36) (120 s.) (3/5 s.) (6 s.) (1/3/6 s.) (6 s.) (3 s.) (9 s.)

Fig. 2.1: Graphical representation of the experimental design. The design consisted of two phases: “Single picture phase” and “Transient pictures phase”. pictures was based on the normative ratings of IAPS along the dimensions of affective valence and arousal (both scales ranging from 1 to 9, i.e. from low arousal to high arousal, and low pleasure to high pleasure). The picture selection was done to create ﬁve different picture classes (Fig. 2.2): • High-arousal-pleasant (HAP) class with mean valence of 7.06 (SD=0.45) and mean arousal of 6.47 (SD=0.34) • Low-arousal-pleasant (LAP) class with mean valence of 7.25 (SD=0.47) and mean arousal of 3.09 (SD=0.29) • High-arousal-unpleasant (HAU) class with mean valence of 1.97 (SD=0.69) and mean arousal of 6.75 (SD=0.30) • Low-arousal-unpleasant (LAU) class with mean valence of 3.57 (SD=0.61) and mean arousal of 3.77 (SD=0.40) • Neutral (N) class with mean valence of 4.54 (SD=0.20) and mean arousal of 4.45 (SD=0.30)

Each class contained twenty-four pictures, except for the neutral class that only included eighteen pictures.

Stimuli for the “Transient picture phase” The formal composition of the transient stimuli consisted of the sequential presentation of two pictures belonging to different picture classes, separated by a blackout interval of a duration of 1, 3, or 6 seconds. The transient stimuli were organized in twelve different classes, created to cover different transitions along the pleasant-to-unpleasant and unpleasant-to-pleasant axis. Tran- 2. Methods 32

Arousal

HAP HAU

N Valence

LAP LAU

Fig. 2.2: Schematic representation of the affective picture classes created for this experiment. Horizontal axis corresponds to valence, vertical axis corresponds to arousal. HAU = high-arousal-unpleasant, HAP = high-arousal-pleasant, LAP = low-arousal-pleasant, LAU = low-arousal-unpleasant. sient stimuli of pleasant-to-unpleasant valence content included transitions of HAP- to-HAU, HAP-to-LAU, HAP-to-Neu, LAP-to-HAU, LAP-to-LAU and LAP-to-Neu. Conversely, transient stimuli of unpleasant-to-pleasant valence content included transitions of HAU-to-HAP, HAU-LAP, HAU-to-Neu, LAU-to-HAP, LAU-LAP and LAU- to-Neu (Fig. 2.3).

2.3 Sample

The sample of the study comprises two groups of participants: Migraineurs with aura (MA), and a control group (Ctr). Eighteen female participants were recruited from the migraineurs’ data base of Hospital Vall D’Hebron to be included in the MA group. Migraineurs included in this study had been previously diagnosed as suffering from migraine with aura, according to the ICHD II criteria (Olesen & Steiner, 2004). Co- morbidity of psychiatric diseases, as coded by the DSM-IV, including depression and anxiety disorders, as well as, analgesic or other substance abuse, were an exclusion criteria. Due to the diversity of preventive treatment approaches (see Appendix A) and the relatively small pool of migraineurs in the Hospitals’ database, differences in treatment approaches were not used as a criteria for patients selection. The selected migraine patients reported a mean of 6.78 (SD=3.66) days with migraine monthly frequency, computed in the last six months. Patients included in the study were in inter-attack phase. In average, resolution of the last migraine crisis was 12.28 (SD=9.79) days before the study, and patients were not experiencing predorme symptoms, as ascertained by interview. Acute treatment drugs were not used by the MA participants in the 24 hours before the study. As a matched control group twenty-two female non-migraineurs 2. Methods 33

Arousal Arousal

HAP HAU HAP HAU

N Valence N Valence

LAP LAU LAP LAU

(a) (b) Arousal Arousal

HAP HAU HAP HAU

N Valence N Valence

LAP LAU LAP LAU

Fig. 2.3: Transient classes created for this experiment. Horizontal axis correspondes to valence, vertical axis corresponds to arousal. Two affective pictures presented sequential separated by an interval of variable duration (1,3 or 6 seconds). HAU= high- arousal-unpleasant, HAP = high-arousal-pleasant, LAP = low-arousal-pleasant, LAU = low-arousal-unpleasant. were recruited locally in Barcelona. Exclusion criteria for the control group were having suffered from migraine episodes or having a family history of migraine. MA and control group were matched for age (MA group: M=39.28 SD=5.85; Ctr group: M=35.24, SD=5.04), nationality (MA group: 94.44% spanish and 5.56% colombian; Ctr group: 86.36% spanish, 9.09% italian and 4.5% colombian), and education (MA group: 50% higher, 18.18% secondary, 13.64% primary; Ctr group: 86.36% higher, 9.09% secondary, 4.54% primary).

2.4 Procedure

Experiments were conducted in a quiet and dark room with controlled temperature. After the subject arrived to the experimental set the experimenter briefed her about the aim of the study, the tasks that will be required and the duration of the experiment. The 2. Methods 34 subject was asked to answer some informative questions, and to sign the legal consent form (see Appendix B). Subsequently, the experimenter applied the physiology elec- trodes and explained to the subject that there will be two phases during the experiment. The subject was seated as comfortably as possible in front of the computer. The subject was then told that during the ﬁrst phase pictures differing in emotion content would be displayed, and that each picture should be attended to during the entire time it was displayed on the screen. The subject was instructed on how to use the rating scales shown on the computer, and the experimenter explained that after the picture offset the subject would be asked to rate how she felt using this instrument. The experimenter informed the subject that the beginning of the second phase of the experiment is marked by a black screen of two minutes. The subject was asked to relax during this period. The subject was told that after this interval, images will be presented again, but this time in sequences of two, separated by a short blackout. The experimenter explained to the subject that after the second image was presented, the subject will be again asked to rate how she felt. The subject was explicitly instructed that she should rate how she felt after seeing both images. At the end of this phase, the experimenter removed the elec- trodes from the participant, and asked her to ﬁll the Behavioral Inhibition Behavioral Activation questionnaire (Carver & White, 1994) (see Appendix C). The experimental protocol was approved by the ethics committee of the Vall D’Hebron Hospital and an informed consent was obtained for all participants (see Appendix B).

2.5 Apparatus

In this section I will describe the hardware and software components used in the implementation of the experiment. Hardware components comprised two networked computers, a “USB.amp” physiology recording kit (Guger Technologies OG, Graz, Austria) with corresponding sen- sors for recording of electrodermal activity, electrocardiogram, respirations and facial electromyography, and sound isolation headphones. Stimuli where presented on a 19" monitor and ratings were given using the computer mouse. The Affective Multimodal Data Base (AMDB) software was used for generating the experimental trial order and randomization, stimuli presentation, synchronization of stimuli presentation with physiological recording and for the gathering and storage of subjective responses. Next, I will brieﬂy describe the design and the development of the AMDB software, and report on the validation in a ﬁrst test use.

2.5.1 Development of the Affective Multimodal Data Base (AMDB) system Initially, the emphasis of the present master project was on the development of the Affective Multimodal Data Base (AMDB) system, which was in an early stage of development at the SPECS laboratory. Though the focus of the master project shifted 2. Methods 35 from the development of the AMDB software to the implementation of an experimental study, signiﬁcant contributions were made to the development of the AMDB. The target functionality of the AMDB system is the interactive composition and presentation of affective stimulation in different modalities, i.e. via pictures, sound and video. Contributions to the development of the AMDB include the addition of a graphical user interface for the recording of subjective ratings (Fig. 2.4), implementation of syn- chronous presentation of stimuli and recording of physiological signals, improvement of the bi-directional communication between the controller and database modules, and the realization of the ﬁrst test case of the AMDB system.

Fig. 2.4: Graphical user interface of the AMDB

System architecture overview The goal functionality of the AMDB was the composition and presentation of affective stimuli in different modalities. This means that the system must be able to display videos and images, and playback sounds in prediﬁned sequences. Next to this, existing ratings for stimulus material, e.g. IAPS, has to be read into the system, and data recorded has to be synchronized with the stimulus with which it was presented. Addi- tionally, the AMDB was targeted to be cross-platform, i.e. to run on different operating systems such as Linux, and Windows. In the development of the AMDB I followed a modular design principle. This approach was choosen so that I could make maximal use of existing components and increase ﬂexibility and extensibility of the system. The AMDB prototype is composed by four main components (Fig. 2.5).: 2. Methods 36

1. The Audiovisual player developed based on the Open Source Computer Vision library (OpenCv, http://sourceforge.net/projects/opencvlibrary/), and the irrK- lang sound engine (Ambiera e.U., Vienna, Austria) 2. The Controller developed in Java and integrating a Graphical User Interface 3. The Physiology recording module developed in Simulink (MathWorks, Natick, Massachusetts, U.S.A.) 4. The relational database implemented in SQLite (http://www.sqlite.org/). The controller is the core module. Its main function is to parse subjective and physiological information from the user and transpose it to database queries that allow the selection and composition of user-directed emotion stimulation that, in turn, is presented through the player module (Fig. 2.5). The relational data base design created for the AMDB system allows matching data by using common characteristics found within the data set. The concept behind the data structure is that the AMDB system crosses user related information with stimuli related information.

Fig. 2.5: Design for the architecture of the AMDB prototype. Scheme depicts information ﬂow. Green arrows indicate information input, whereas blue arrows indicate information output

Validating the AMDB with a test case The ﬁrst test case for the AMDB prototype was the use in the compilation of affective video library for the SPECS lab. To compile a experimentally validated affective video library, 50 video clips of 10 seconds each, were edited from raw footage material produced by the video artist Behdad Rezazadeh. The AMDB system was the software 2. Methods 37 supporting the process of video presentation and rating. Subjective ratings were done using the Self-Assessment-Manikin (SAM), a non-verbal picture based self assessment questionnaire that measures the dimensions pleasure, arousal, and dominance (P. Lang et al., 2008). A digital version of the SAM, reduced to the dimension of pleasure and arousal, was integrated into the AMDB’s graphical user interface. A total of twenty participants (10 female and 10 male) with ages ranging from 20 to 35 rated the 50 video clips (Fig. 2.6). 8 6 Valence 4 2

2 4 6 8 Arousal

Fig. 2.6: Scatterplot of the ratings distribution of the video clips in the valence-arousal space 2. Methods 38 3. RESULTS

Data collected during the experimental phases of the empirical study conducted consisted on subjective ratings, physiology responses and Behavioral Inhibition Behav- ioral Activation questionnaires. Analysis of the subjective ratings were included in the scope of this work. In the following, I will present the results obtained and describe the statistical methods used.

3.1 Affective responses to single pictures

The first step performed in the analysis of the data collected from the first phase of the study (“Single picture phase”) was a validation of the affective pictures classification system used (High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neu), Low-Arousal-Pleasant (LAP) and Low-Arousal-Unpleasant (LAU)). Following that, I assessed IAPS affective induction predictions for the control group and the migraineurs with aura (MA) group. Furthermore, I compared the affective responses to picture classes between both groups. Finally, an analysis of the effects in repeated exposure to affective stimuli was conducted.

3.1.1 Validation of the affective pictures classification system The selection of affective stimuli for this study was done to achieve three distinct valence classes – Pleasant, Neutral and Unpleasant; and three distinct arousal classes – High, Neutral and Low. I examined the robustness of this classification by probing the differentiation of subjective ratings between classes. For each participant group, Control and MA, comparisons using Wilcoxon Signed Ranks test with Bonferroni correction were conducted. Comparisons between valence classes, showed significant differences between Pleasant and Neutral stimuli [Control group: r=0.54, p< 0.001; MA group: r= 0.57, p<0.001], between Neutral and Unpleasant stimuli [Control group: r= 0.31, p<0.001; MA group: r= 0.34, p<0.001] and between Pleasant and Unpleasant stimuli [Control group: r=0.73, p<0.001; MA group: r= 0.67, p<0.001] (Fig. 3.1a). In addition, comparisons between arousal classes, showed significant differences between High and Neutral stimuli [Control group: r=0.16, p<0.01; MA group: r= 0.25, p<0.001], between Neutral and Low stimuli[Control group: r= 0.35, p<0.001 ;MA group: r= 0.13, p<0.05], and between High and Low stimuli [Control group: r=0.52, p<0.001; MA group: r= 0.36, p<0.001](Fig. 3.1b). 3. Results 40

Distinction of valence classes Distinction of arousal classes 7 6 6 5 5 4 4 3 3 2 Arousal ratinngs 2 Valence ratinngs Valence 1 1 0 0 Ctr.P MA.P Ctr.N MA.N Ctr.U MA.U Ctr.H MA.H Ctr.N MA.N Ctr.L MA.L Group . Valence class Group . Arousal class

(a) (b)

Fig. 3.1: Differentiation between picture classes for the Control group and the MA group. a) Participant groups are Control (Ctr) and migraineurs with aura (MA). Valence types are Pleasant (P), Neutral (N) and Unpleasant(U). b) Participant groups are Control (Ctr) and migraineurs with aura (MA). Arousal types are High (H), Neutral (N) and Low(L). Both groups successfully differentiated between arousal classes and valence classes, supporting the affective picture classiﬁcation used in this experiment.

Measures of central tendency and dispersion were computed to summarize the subjective ratings data. As expected, mean valence ratings decrease from Pleasant to Neu- tral, to Unpleasant classes. Mean arousal ratings decrease from High to Neutral, to Low classes (Table 3.1).

Tab. 3.1: Mean scores for valence and arousal picture classes.

Control group1 MA group Valence Types Pleasant 6.81 (1.61) 7.19 (1.87) Neutral 4.31 (2.06) 3.98 (2.02) Unpleasant 2.99 (1.87) 2.53 (2.23) Arousal Types High 6.33 (1.72) 6.67 (2.21) Neutral 5.65 (1.96) 5.43 (2.22) Low 4.26 (1.88) 4.71 (2.45)

These results support the classiﬁcation of affective stimuli proposed for this experiment (High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neu), Low-Arousal-Pleasant (LAP) and Low-Arousal-Unpleasant (LAU)) by show- 3. Results 41 ing that pictures classiﬁed in the valence classes Pleasant, Neutral and Unpleasant and in the arousal classes High, Neutral, Low are distinguished by the subjective ratings produced by participants of both groups.

3.1.2 Assessment of IAPS affective induction predictions The affective pictures used in this experiment were selected from the IAPS library. IAPS is a normalized collection of affective stimuli where each picture is categorized by a mean and standard deviation score for valence, arousal and dominance. In the subsequent analysis I examine if IAPS affective induction predictions were maintained for both groups in the study. Linear regressions were computed to assess the relationship between IAPS values and the ratings of each group. IAPS valence scores significantly predicted the valence ratings of both groups [Control: b = 0.83, t(28) = 11.45, p < 0.001 ; MA: b = 0.98, t(28) = 12.51, p < 0.001]. In both groups, the linear regression model was able to explain a large amount of the variance in the valence scores [Control: adjusted R2= 0.82, F(1, 28) = 131.2, p < 0.001 ;MA: adjusted R2= 0.84, F(1, 28) = 156.4, p < 0.001] (Fig. 3.2a and Fig. 3.2b). For the arousal dimension, IAPS arousal values significantly predicted the arousal ratings of both groups [Control: b = 0.70, t(28) = 8.91, p < 0.001; MA: b = 0.75, t(28) = 5.77, p < 0.001 ]. However the linear regression model computed for the MA group (Fig. 3.3b) was less suitable to explain the variation in the arousal data than the one compute for the Control group (Fig. 3.3a) [Control: adjusted R2= 0.73, F(1, 28) = 79.4, p < 0.001; MA: adjusted R2= 0.52, F(1, 28) = 32.29, p < 0.001]. IAPS values significantly predicted ratings on both groups. Increases of IAPS values were always correlated with increases in subjective ratings. However the lower performance of IAPS in predicting the variances in the arousal ratings from the MA group, points to possible differences in migraineurs reactivity to arousing content.

3.1.3 Comparison between migraineurs and non-migraineurs responses The aim of the ﬁrst phase of the experimental design of this study (“Single Picture Phase”) was to probe group differences in affective responses to the different picture classes, as well as to provide baseline data for the second phase of the experiment (“Transient Pictures Phase”). As mentioned before, the picture classes created for this study are: High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neu), Low-Arousal-Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). To determine if affective responses to picture classes varied between the Control and the MA group the mixed model ANOVA method was used, to assess the impact of picture class on the participants’ ratings. The mixed model ANOVA for the mean valence ratings showed a main effect of picture class [F(4, 148)=252.54, p<0.001] and an interaction effect between picture class and group [f(4,148)=5.49, p<0.001] (Fig. 3.4). Post-hoc analysis of the interaction effect showed a highly signiﬁcant difference between groups for mean valence ratings of LAU pictures [F(1,37)=11.06, p<0.001]. 3. Results 42

IAPS accuracy − Valence (Control group) IAPS accuracy − Valence (MA group) 8 8 7 7 6 6 5 5 4 4 3 3 2 2 Mean valence ratings (MA group) ratings Mean valence Mean valence ratings (Control group) ratings Mean valence 2 3 4 5 6 7 8 2 3 4 5 6 7 8 IAPS values IAPS values

(a) (b)

Fig. 3.2: Accuracy of IAPS valence induction predictions for Control and MA group. a) Scat- terplot assessing the relationship between mean valence ratings from the Control group and IAPS valence scores. b) Scatterplot assessing the relationship between mean valence ratings from the MA group and IAPS valence scores. IAPS values sig- niﬁcantly predicted valence ratings on both groups. Increases of IAPS valence values were always correlated with increases in valence ratings.

Tab. 3.2: Valence and arousal mean scores for each picture class

Valence Ratings2 Arousal Ratings Picture Types Control Group MA Group Control Group MA Group HAP 7.14 (1.50) 7.30 (1.81) 6.48 (1.60) 6.37 (1.99) LAP 6.48 (1.66) 7.08 (1.94) 3.93 (1.83) 3.73 (2.34) Neutral 4.31 (2.06) 3.98 (2.02) 5.65 (1.96) 5.43 (2.22) LAU 3.52 (1.93) 2.45 (2.07) 4.59 (1.87) 5.69 (2.17) HAU 2.47 (1.64) 2.61 (2.39) 6.19 (1.82) 6.97 (2.39)

Mean valence ratings of LAP pictures achieved a nearly significant difference between groups [F(1,37)=3.36, p=0.07]. Whereas no significant differences between groups where found for mean valence ratings of HAP, HAU and Neutral pictures. The mixed model ANOVAfor the arousal ratings showed a main effect of picture class [F(4, 148)= 38.88, p<0.001] and an interaction effect between picture class and group [f(4,148)= 3.06, p=0.01]. Post-hoc analysis on the interaction effect showed significant differences between groups for mean arousal ratings of LAU [F(1,37)=10.24, p<0.01] and HAU pictures [F(1,37)= 4.073, p<0.05], whereas the mean arousal ratings of HAP, LAP and Neutral pictures were not significant different between groups. These results point to a higher susceptibility of migraineurs to negative stimuli, as 3. Results 43

IAPS accuracy − Arousal (Control group) IAPS accuracy − Arousal (MA group) 7.0 6.5 7 6.0 6 5.5 5.0 5 4.5 4 4.0 3 3.5 Mean arousal ratings (MA group) Mean arousal ratings Mean arousal ratings (Control group) Mean arousal ratings 3 4 5 6 7 3 4 5 6 7 IAPS values IAPS values

(a) (b)

Fig. 3.3: Accuracy of IAPS arousal induction predictions for Control and MA group. a) Scat- terplot assessing the relationship between mean arousal ratings from the Control group and IAPS arousal scores. b) Scatterplot assessing the relationship between mean arousal ratings from the MA group and IAPS arousal scores. IAPS values sig- niﬁcantly predicted arousal ratings on both groups. Nonetheless, the linear regression model computed for the MA group performed worse in predicting variances in the arousal scores.

Mean valence ratings (Control and MA group)

Group 7

Control CMA 6 5 4 Valence ratings Valence 3

HAP HAU LAP LAU Neut Picture class

Fig. 3.4: Valence affective responses to picture types for Control and MA group. Picture classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). Pictures of class LAU were rated by migraineurs as more unpleasant. 3. Results 44

Mean arousal ratings (Control and MA group) 7.0 Group

Control 6.5 CMA 6.0 5.5 5.0 Arousal ratings 4.5 4.0

HAP HAU LAP LAU Neut Picture class

Fig. 3.5: Arousal affective responses to picture types for Control and MA group. Picture classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). Unpleas- ant stimuli were considered signiﬁcantly more arousing by migraineurs then by non- migraineurs. unpleasant visual content was considered more arousing by migraineurs than by non- migraineurs. Furthermore, migraineurs reported more negative affective states after viewing Low-Arousal-Unpleasant pictures than non-migraineurs.

3.1.4 Analysis of the effects in repeated exposure to affective stimuli To examine if the repeated exposure to affective stimuli produced differences in the subjective responses, Pearson correlation coefficients assessing the relationship between the stimuli order of presentation and the subjective ratings were computed. For this analysis I arranged the data in three valence related groups - Unpleasant (joining classes HAU and LAU), Neutral (corresponding to the Neutral class) and Pleasant (joining classes HAP and LAP); and two arousal related groups - High (joining classes HAU and HAP) and Low (joining classes LAU and LAP). For Unpleasant, Neutral and Pleasant pictures, correlations between order of presentation and subjective ratings (valence and arousal) were computed. Additionally, for High and Low pictures correlations between order of stimuli presentation and arousal ratings were computed. A significant negative correlation between the valence ratings of Neutral pictures and their order of presentation was found for the MA group [Pearson’s r(27) = -0.36, p < 0.05] (Fig. 3.6b) , but no significant correlation was found for the Control group [Pearson’s r(27)=-0.29, p=0.12] (Fig. 3.6a). Furthermore, only the MA group revealed a significant negative correlation between the valence ratings of Unpleasant pictures and their presentation order [MA: Pearson’s r(27) = -0.45, p < 0.05; Control: Pear- 3. Results 45 son’s r(27)=1.82, p=0.09] (Fig. 3.7a and Fig. 3.7b). For both groups, no significant correlations were found neither for the valence ratings of Pleasant pictures or for the arousal dimension of Pleasant, Neutral and Unpleasant picture groups. Correlational analysis for the arousal categories (High and Low) didn’t reveal any significant correlations between the order of presentation of the stimuli and the arousal ratings.

Control group MA group 6.0 6 5.5 5 5.0 4.5 4 4.0 3 3.5 Mean valence rating Mean valence Mean valence ratings Mean valence 3.0 2

0 5 10 15 20 25 30 0 5 10 15 20 25 30 Order of presentation Order of presentation

(a) (b)

Fig. 3.6: Relationship between order of presentation and valence scores for Neutral pictures. a) Presents the scatterplot for the Control group, where presentation order was found not to be signiﬁcantly correlated to the affective experience of the neutral content. b) Presents the scatterplot for the MA group, where Neutral pictures’ presentation order and valence ratings were negatively correlated. Denoting that the more neutral pictures participants of the MA group viewed the less pleasant they rated their affective experience of those pictures.

These results show that for the Control group, stimuli presentation order had no signiﬁcant repercussions on valence and arousal ratings. Whereas for the migraineurs group, the more times a participant was exposed to neutral or to unpleasant content the more negatively he reported his affective experience. Therefore, suggesting that migraineurs experience an affective accumulation effect when subjected to repeated non-pleasant content.

3.2 Affective responses to transient stimuli

In the analysis of the data collected from the second phase of the study (“Transient picture phase”) I inspected the existence of the valence inhibition effect in both groups. Afterward, I compared the amplitude of this effect between both groups in transitions from pleasant to unpleasant content and transitions from unpleasant to pleasant content. In parallel, I assessed the arousal modulation that accompanied the valence 3. Results 46

Control group MA group 6 4.0 5 3.5 4 3.0 3 2.5 2 Mean valence ratings Mean valence ratings Mean valence 2.0 1 1.5

0 5 10 15 20 25 30 0 5 10 15 20 25 30 Order of presentation Order of presentation

(a) (b)

Fig. 3.7: Relationship between order of presentation and valence scores for Unpleasant pictures. a) Presents the scatterplot for the Control group, where the presentation order was found not to be signiﬁcantly correlated to the affective experience of unpleasant stimuli. b) Presents the scatterplot for the MA group, where Unpleasant pictures’ presentation order and valence ratings were negatively correlated. Revealing that the more unpleasant pictures participants of the MA group viewed the more unpleasant they rated their subjective experience of those pictures. inhibition effect. To further, examine the predictions of the valence inhibition effect I searched for correlations between the arousal power of the second picture presented and the amplitude of the valence inhibition. Moreover, I probed the temporal modulation of valence inhibition and arousal modulation in transient stimuli.

3.2.1 Analysis of the Valence Inhibition effect To study the valence inhibition effect - the phenomenon that refers to alterations in the valence experience when two pictures of contradictory valence are presented in sequence, where the arousal power of the second picture modulates the valence subjective state induced by the first picture (Schimmack, 2007) - I considered the first picture presented as the context, setting an initial affective state, and the second picture presented as the inhibitor, modulating the affective state induced before. The pictures presented, as context and inhibitor, in the second phase of the experiment, were not the same pictures presented in the first phase. Still they were selected from the same valence/arousal clusters that defined the classes of HAU, LAU, Neutral, HAP and LAP used in the first phase. This way I avoided pre-exposure effects that would arise from using the same pictures on both phases, while securing continuity between the picture classes of both phases. To measure the amplitude of the valence inhibition effect, I need to assess the affec- 3. Results 47 tive state of the participant after the context stimuli presentation, in order to compare it to the affective state achieved after the inhibitors’ presentation. Since stimuli transitions cannot be interrupted by a rating I used the data collected in the first phase of the experiment to estimate each participants’ affective state after the context picture of each trial. This was done by computing linear regression models based on the data from the first phase, relating the IAPS scores of the pictures presented in the first phase with the correspondent valence and arousal ratings of each participant. In this way I were able to generate predictive models of affective responses to IAPS pictures for each participant individually. These predictive models were then used to estimate, based on the IAPS scores of context pictures, the affective state induced in each participant by the context picture presentation. Based on this I computed the amplitude of the valence inhibition effect for each participants’ trial, by subtracting the estimated valence state induced by the context picture from the valence rating done at the end of the transient stimuli presentation. Similarly, I computed the arousal modulation effect by subtracting the estimated arousal state induced by the context picture from the arousal rating of each transient stimuli trial. To analyze the valence inhibition effect I grouped the transient stimuli data in two categories: Pleasant-To-Unpleasant (joining data from stimuli transitions starting with pictures from HAP and LAP classes, including all interval variations of 1 second, 3 seconds and 6 seconds); and Unpleasant-To-Pleasant (joining data from transitions starting with pictures from HAU and LAU classes, including all interval variations of 1 second, 3 seconds and 6 seconds).

Transitions from Pleasant to Unpleasant Firstly, I probed the existence of the valence inhibition effect within each group, by conducting comparisons using the Friedman test to assess differences in ratings of Pleasant-To-HAU, Pleasant-To-Neutral and Pleasant-To-LAU transient stimuli. As expected, this analysis revealed a valence inhibition effect for both groups. As, valence ratings of transient stimuli with the pleasant context showed to be differently modulated by the arousing power of different inhibitor classes [Control (Pleasant to Unpleasant): χ2(2)=97.27, p < 0.001; MA (Pleasant to Unpleasant): χ2(2)=40.98, p < 0.001]. Afterward, I compared the amplitude of the valence inhibition effects between both groups using the Mann-Whitney test with Bonferroni correction. This analysis showed that the valence inhibition observed in Pleasant-To-HAU, Pleasant-To- Neutral and Pleasant-To-LAU stimuli was significantly different between the Control and the MA group. [Pleasant-To-HAU: r=2.40, p<0.05; Pleasant–To-Neutral: r=2.37, p<0.05; Pleasant–To-LAU: r=3.10, p<0.01] (Fig. 3.8). To examine if the valence inhibition effect was accompanied by differences in the subjective experience of arousal, I applied the same statistical method to the arousal dimension. This revealed that the arousal modulation accomplished by the inhibitor was significantly different between both groups for Positive-To-HAU transitions [r=-2.54, p<0.05], whereas the 3. Results 48 arousal modulation of Positive-To-Neutral transitions only reached nearly significant differences [r=2.32, p=0.06], and Positive-To-LAU transitions revealed to be not significantly different between both groups (Fig. 3.9).

Valence Inhibition Effect (Pleasant−To−Unpleasant) 0 −1 −2 −3 Valence Inhibition Effect Valence −4

Ctr.HAU MA.HAU Ctr.LAU MA.LAU Ctr.Neu MA.Neu Group . Inhibitor class

Fig. 3.8: Valence Inhibition effect for Pleasant to Unpleasant transient stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal- Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal- Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). Valence inhibition amplitudes of transitions starting from a pleasant context were found to be signiﬁcantly different between groups [Pleasant-HAU: r=2.40, p<0.05; Pleasant – LAU: r=3.10, p<0.01; Pleasant – Neutral: r=2.37, p<0.05].

The valence inhibition effect is grounded on the impact of the inhibitors’ arousal in the valence state induced by the context. To further inspect this relationship linear regressions to assess the relationship of inhibitors’ arousal and inhibition effect were computed. For both groups the amplitude of the valence inhibition effect was found to be signiﬁcantly related with the estimated arousal of the inhibitor picture [Control: b = -0.99, t(378) = 1.908, p < 0.001; MA: b = -0.49, t(378) = 2.64, p < 0.001]. However the linear regression model for the Control group is able to explain approximately 30% of the variations on the amplitude of the valence inhibition effect (Fig. 3.10a), whereas the model for the MA group only explains up to 10% of the variance in valence inhibition data [Control: adjusted R2= 0.3, F(1, 378) = 124.3 , p < 0.001; MA: adjusted R2= 0.1 , F(1, 378) = 22.69, p < 0.001](Fig. 3.10b).

Transitions from Unpleasant to Pleasant To probe the existence of the valence inhibition effect within each group, comparisons using the Friedman test were conducted, to assess differences in ratings of Unpleasant- To-HAP, Unpleasant-To-Neutral and Unpleasant-To-LAP stimuli. This analysis re- 3. Results 49

Arousal modulation (Pleasant−To−Unpleasant) 1.5 1.0 0.5 0.0 Arousal modulation −0.5

Ctr.HAU MA.HAU Ctr.LAU MA.LAU Ctr.Neu MA.Neu Group . Inhibitor class

Fig. 3.9: Arousal modulation accompanying the valence inhibition effect for Pleasant to Un- pleasant stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). Signifi- cantly different arousal modulation between groups was found only for transitions from Positive to HAU [r=-2.54, p<0.05]. vealed a valence inhibition effect for both groups. As expected, valence ratings of transient stimuli with unpleasant context showed to be differently modulated by the arousing power of different inhibitor classes [Control (Unpleasant to Pleasant): χ2(2)=41.15, p < 0.001; MA (Unpleasant to Pleasant): χ2(2)=27.72 , p < 0.001]. Comparisons of the amplitude of the valence inhibition effect in Unpleasant-To-Pleasant stimuli were performed with the same methods described above for the Pleasant-To-Unpleasant stimuli. This analysis revealed no significant differences of valence inhibition amplitude (Fig. 3.11) or arousal modulation (Fig. 3.12) between groups. As for the Pleasant-To-Unpleasant transition analysis, linear regressions were computed to further inspect the relationship between the arousal of the inhibitor picture and the amplitude of the valence inhibition effect in Unpleasant-To-Pleasant stimuli. For both groups the amplitude of the valence inhibition effect was significantly related to the estimated arousal of the inhibitor picture presented [Control: b = -0.49, t(378) = 2.64, p < 0.001; MA: b = -0.28, t(378) = -2.82, p < 0.001] (Fig. 3.13). Though for Unpleasant-To-Pleasant stimuli, the linear regression model for the Control group is only able to explain approximately 10% of the variations of the amplitude of the valence inhibition effect, while the model for the MA group explains no more than 2% of the variance in the data [Control: adjusted R2= 0.1 , F(1, 378) = 22.69, p < 0.001; MA: adjusted R2= 0.02 , F(1, 378) = 7.95, p < 0.001]. On the one hand these results showed that the migraineurs group was more sus- ceptible to negative valence changes induced by non-pleasant content after positive af- 3. Results 50

Control group MA group 4 2 2 0 0 −2 −2 −4 −4 Valence Inhibition Valence Inhibition Valence −6 −6 −8 3 4 5 6 7 8 2 3 4 5 6 7 8 9 Inhibitors' arousal Inhibitors' arousal

(a) (b)

Fig. 3.10: Relationship between the arousal of the inhibitor picture and the valence inhibition effect for Pleasant-To-Unpleasant stimuli. a) For the Control group a signiﬁcant correlation was found, explaining 30% of the valence inhibition data variation [Control : b = -0.99, t(378) = 1.908, p < 0.001, adjusted R2= 0.3 , F(1, 378) = 124.3]. b) For the MA group, a signiﬁcant correlation found explaining 10% of the valence inhibition data variation [MA: b = -0.49, t(378) = 2.64, p < 0.001, adjusted R2= 0.1 , F(1, 378) = 22.69, p < 0.001]. fective induction (additionally in Pleasant-To-HAU transitions, the migraineurs group experienced a higher arousal modulation). On the other hand, migraineurs and controls were equally robust on recovering from a negative induced state after the presentation of pleasant content (too no differences in the arousal modulation were found between groups in Unpleasant-To-Pleasant transitions). Correlations between the estimated inhibitors’ arousal and the amplitude of the valence inhibition for Pleasant- To-Unpleasant and Unpleasant-To-Pleasant transitions showed that the expectations postulated by the valence inhibition effect were better observed in the Control group than in the MA group, pointing to a lower resolution response to valence modulation by arousal in the migraineurs group.

3.2.2 Probing the temporal modulation of Valence Inhibition and arousal modulation The experimental design of this study included variations of the time interval separating the transient pictures (1, 3 or 6 seconds). To probe whether this interval differences had signiﬁcant consequences for the amplitude of the valence inhibition and accompanying arousal modulation, a Man-Whitney test was applied, comparing data from trials with different time intervals for each transient stimuli class within each group (e.g. comparison between the valence inhibition amplitude in the MA group between 3. Results 51

Valence Inhibition Effect (Unpleasant−To−Pleasant) 2.0 1.5 1.0 0.5 Valence Inhibition Effect Valence 0.0 Ctr.HAP MA.HAP Ctr.LAP MA.LAP Ctr.Neu MA.Neu Group . Inhibitor class

Fig. 3.11: Valence Inhibition effect for Unpleasant to Pleasant transient stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal- Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal- Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). Valence inhibition amplitudes of transitions starting from a unpleasant context stimuli were found to be similar between both groups. conditions of HAP-To-HAU with a 1 second interval, 3 seconds interval and 6 seconds interval). For both groups, this analysis revealed that for all transition classes the amplitude of the valence inhibition (Fig. 3.14) and arousal modulation (Fig. 3.15) was not signiﬁcantly different between different interval conditions. Hence, time intervals of 1, 3 or 6 seconds had no signiﬁcant repercussions on the amplitude of the valence inhibition and accompanying arousal modulation. 3. Results 52

Arousal Modulation (Unpleasant−To−Pleasant) 0.5 0.0 −0.5 Arousal modulation −1.0 Ctr.HAP MA.HAP Ctr.LAP MA.LAP Ctr.Neu MA.Neu Group . Inhibitor class

Fig. 3.12: Arousal modulation accompanying the valence inhibition effect for Unpleasant to Pleasant transient stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal-Pleasant (HAP), High-Arousal- Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal- Unpleasant (LAU). Arousal modulation found in Unpleasant-To-Pleasant stimuli was similar between both groups.

Control group MA group 8 8 6 6 4 4 2 2 0 0 Valence inhibition effect Valence inhibition effect Valence −2 −2

2 3 4 5 6 7 8 2 3 4 5 6 7 8 9 Inhibitors' arousal Inhibitors' arousal

(a) (b)

Fig. 3.13: Relationship between the arousal of the inhibitor picture and the valence inhibition effect Unpleasant-To-Pleasant transitions. a) For the Control group, a signiﬁcant correlation was found explaining 10% of the valence inhibition data variation [b = -0.49, t(378) = 2.64, p < 0.001, adjusted R2= 0.1 , F(1, 378) = 22.69, p < 0.001]. b) For the MA group a signiﬁcant correlation was found, explaining 2% of the valence inhibition data variation [b = -0.28, t(378) = -2.82, p < 0.001, adjusted R2= 0.02 , F(1, 378) = 7.95, p < 0.001]. 3. Results 53

HAU−To−HAP HAP−To−HAU 0 0 −2 −2 −4 −4

1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA HAU−To−Neut HAP−To−Neut −0.5 1.0 −2.0 0.0

1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA HAU−To−LAP HAP−To−LAU 0.0 1.5 −1.0 0.0 −2.5 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA LAU−To−HAP LAP−To−HAU −1 2.0 −3 1.0 −5 0.0 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA LAU−To−Neut LAP−To−Neut −0.5 0.5 −0.5 −2.0

1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA LAU−To−LAP LAP−To−LAU 0.0 2.0 −1.5 1.0 0.0 −3.5 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA

Fig. 3.14: Valence Inhibition in transitions with different intervals. White bars corresponded to the Control group (1, 2 and 3 seconds correspond to Ctr.1, Ctr.3 and Ctr.6). Filled bars corresponded the MA group (1, 2 and 3 seconds are MA.1, MA.3 and MA.6). Vertical axis indicates the amplitude of the valence inhibition effect. 3. Results 54

HAU−To−HAP HAP−To−HAU 0.5 0.5 −0.5 −0.5

1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA HAU−To−Neut HAP−To−Neut −0.5 −1.0 −1.5 −2.5 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA HAU−To−LAP HAP−To−LAU −0.5 −0.5 −1.5 −2.0

1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA LAU−To−HAP LAP−To−HAU 2.0 3.0 1.0 1.5 0.0 0.0 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA LAU−To−Neut LAP−To−Neut 0.5 0.5 −0.5 −0.5 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA LAU−To−LAP LAP−To−LAU 1.0 −0.2 0.0 −1.0

1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA 1.Ctr 3.Ctr 6.Ctr 1.MA 3.MA 6.MA

Fig. 3.15: Arousal modulation in transitions with different intervals. White corresponded to the Control group (1, 2 and 3 seconds correspond to Ctr.1, Ctr.3 and Ctr.6). Filled bars corresponded the MA group (1, 2 and 3 seconds are MA.1, MA.3 and MA.6). Vertical axis indicates the arousal modulation. 4. DISCUSSION AND CONCLUSION

In this last section, I will start by summarizing the work developed during the master. Afterward, I will describe the key results found on the empirical study conducted, contextualize the ﬁndings with literature on the topic of migraine and emotion, and address speciﬁc limitations of the paradigm used in the study. Finally, I will propose future directions that may follow the work developed.

4.1 Summary of thesis work

This thesis was developed in the context of the scientific study of emotion, focusing on the affective modulation through perceptual stimuli. Two goals were pursued during the course of this work. Firstly, the development of the Affective Multimodal Data Base (AMDB) system, a tool for multimodal emotion induction, and secondly an empirical study was conducted, using the AMDB, that aimed at probing the modulation of affective response patterns in migraineurs. The work contributed to the development of the AMDB system by adding a graphical user interface dedicated to subjective ratings, implementing synchronized stimuli presentation and physiology response recording, improving the bidirectional communication between the controller and database modules, and realizing of the first test case of the AMDB system. The empirical study conducted was focused on migraine because of the strong association between migraine and emotion mechanisms. Driven by the goal of study- ing the specific affective profile of migraineurs, an experimental study was designed and conducted with the concrete aim of examining migraineurs dynamic patterns of affective modulation in response to shifts of emotion stimulation. Concretely, I assessed and compared the affective responses of eighteen migraineurs with aura (MA) and twenty-two non-migraineurs to individually presented emotion induction stimuli, and to compositions of transient stimuli that shifted from one valence context to the opposite one, i.e., transitions of pleasant-to-unpleasant stimulation and vice-versa.

4.2 Summary of results

In the following, I will give an overview of the key ﬁndings from the ﬁrst phase of the study ‘‘Single picture phase” and second phase of the study ‘‘Transient pictures phase”. 4. Discussion and Conclusion 56

4.2.1 Single stimuli Findings from the ﬁrst phase of the study, showed that IAPS scores were consider- ably less accurate on predicting the arousal ratings of MA than the arousal ratings of non-migraineurs. In contrast to this, IAPS scores were equally accurate on predicting valence ratings for both groups. Higher arousal impact in MA than in controls was elicited by unpleasant pictures (from high-arousal-unpleasant and low-arousal- unpleasant classes). More negative affective states in MA than in controls were elicited by the presentation of unpleasant pictures with low arousal. Conversely, MA and non-migraineurs presented similar affective responses to neutral and pleasant stimuli. Hence, results from the ﬁrst phase of the study revealed that MA had an enhanced susceptibility to negative stimuli in comparison to controls. A potentiation of negative affective states related to the repeated presentation of neutral and unpleasant stimuli was found only for the MA group. Findings revealed that the more times a MA participant was exposed to neutral or to unpleasant content the more negatively she reported her affective experience, whereas previous presentation of stimuli were not related to the affective ratings in non-migraineurs.

4.2.2 Transient stimuli For MA and control group, the presentation of an inhibitor picture (second picture presented) modulated the valence experience of the transient stimuli in accordance to the its’ arousal power. This showed that the valence inhibition effect was observed in both groups. MA have a higher negative valence modulation than controls in responses to transient stimuli of all classes of pleasant-to-unpleasant content, as measured by the amplitude of the valence inhibition effect. Arousal modulation differences between groups, in pleasant to unpleasant stimuli were found in the case of pleasant-to-HAU, but not in pleasant-to-LAU and pleasant-to-neutral transitions. In contrast to this, valence inhibition amplitudes in response to transitions of unpleasant-to-pleasant content, i.e. unpleasant-to-HAP, unpleasant-to-neutral and unpleasant-to-LAP, were similar for MA and controls. In addition, for none of the classes of unpleasant-to-pleasant stimuli, arousal modulation vary significantly between MA and controls. Interval variations of 1, 3 and 6 seconds separating context picture and inhibitor picture had no significant effect, either on the amplitude of valence inhibition nor accompanying arousal modulation. Taken together, the findings from this study suggest an affective profile specific to migraineurs with aura that is characterized by: 1. An enhanced impact of negative stimuli 2. An accumulation of negative affective states when repeatedly exposed to non- pleasant stimuli 4. Discussion and Conclusion 57

3. A high susceptibility to disrupt positive affective states in the presence of unpleasant environmental stimuli

4.3 Contextualization of results

In the following, I will suggest interpretations of the results found in the study, and contextualize the results in the literature on the topic of migraine and emotion. Previous studies indicate, that the IAPS is able to elicit distinctive responses to stimuli in different mental disorders, including mood disorders and schizophrenia (Davidson, 1998; Jayaro, de La Vega, Díaz-Marsá, Montes, & Carrasco, 2008). The less accurate capacity of IAPS for predicting arousal responses of MA may be considered as an indicator of a differentiated emotion processing of arousal which is specific to migraineurs. Findings from this study point to an enhanced response of migraineurs-with-aura to unpleasant stimuli. Contrary to these findings, a previous work using IAPS pictures in a study with migraineurs without aura reported enhanced affective responses to pleasant stimuli, as well as to unpleasant (Tommaso et al., 2009). This difference in results may be due to variation in stimuli selection, experimental procedure or differences in the sample of migraineurs studied (migraineurs-with-aura vs. migraineurs without-aura). In the study here presented, conducted during migraineurs inter-attack phase, a potentiation of negative affective states in MA was observed, as a result of repeated exposure to unpleasant and neutral stimuli. I propose that this potentiation effect is related to cortical non-habituation. Since, several migraine studies using evoked potential methods, have consistently shown that cortical response habituation to perceptual stimuli repetition is decreased in the inter-attack phase of migraine. By contrast, before the attack and during the attack, habituation increases and normalizes (Coppola et al., 2005 cited in Ambrosini, Magis, & Schoenen, 2010). Ambrosini (2010) proposes that electrophysiological changes resulting from habituation are related to a further decrease of serotonergic neurotransmission before a migraine attack, which changes to increased serotonergic transmission during the attack. Due to the prominent role of serotonergic pathways in emotion processing, I hypothesize that the mechanism responsible for serotonergic changes that leads to cortical non-habituation during the inter-attack phase, is involved in the potentiation of negative affective states in response to repeated stimulation, found in the study. Affective modulation via transient affective stimuli showed that migraineurs with aura react more strongly than non-migraineurs to negative modulation of positive affective states in the presence of non-pleasant stimuli. This finding is consistent with the association of migraine with emotion dysfunction expressed in migraineurs comorbidity with depression and anxiety (Torelli & D’Amico, 2004; Radat & Swendsen, 2005; Frediani & Villani, 2007), migraine association with personality traits of neuroticism (K. Merikangas, 1993b; Fan et al., 1999; Bag et al., 2005; Sánchez-Román 4. Discussion and Conclusion 58 et al., 2007; Abbate-Daga et al., 2007; Corr & Matthews, 2009), and the specific role of distress in triggering migraine attacks (Schoonman et al., 2007; Levy et al., 2009; Wober & Wober-Bingol, 2010). Interval duration between the pictures presented in a transition showed no significant difference, which may be due to the high inter-individual variation in affective responses to transient stimuli, that potentially masks small amplitude differences between the affective responses to transitions with different intervals. Although the MA participants selected for this study did not suffer from psychiatric comorbidities personality traits of anxiety and depression are common in migraine pathology (K. Merikangas, 1993b; Fan et al., 1999; Bag et al., 2005; Sánchez- Román et al., 2007; Abbate-Daga et al., 2007; Corr & Matthews, 2009) and might modify emotion responses (Davidson, 2000). In addition, most of the MA participants (72.22%) were under pharmacological treatment that acts on the central nervous system, which may alter their affective state and their responses to the emotion manip- ulations in the study.

4.4 Critical evaluation of the experimental paradigm

Pictures composing the transient stimuli were chosen a priori based on the normalized IAPS values of valence and arousal. Findings from the first phase of the experiment showed differences in affective responses between MA and controls. Hence, it can be assumed that the transient stimuli used, were inherently different in emotion induction between both groups because of affective differences in the individual pictures composing the transition. Consequently, when measuring the amplitude of valence inhibition in response to transient stimuli (in the second phase of the study), it is not possible to fully differentiate whether the effect is solely a result of the modulation achieved through the inhibitor picture, or if it results, in part, from differences in affective responses to the individual pictures composing the transition. I tried to take this limitation into account in the analysis of the data collected. Instead of computing the amplitude of the valence inhibition effect by subtracting the IAPS valence value of the context picture from the valence rating of the transient stimuli, I substituted the IAPS value by an estimate of the valence response of the participant to the context picture. This estimate was computed using a linear regression model based on the responses of each participant to the pictures presented individually in the first phase of the experiment. I assume that this procedure allowed to minimize the limitation of composing transient stimuli with pictures that presumably result in different affective induction outcome between both groups. Nevertheless, the transient stimuli composition remains a methodological limitation of the study. Another limitation related to the transient stimuli composition stems from the fact that the juxtaposition of two pictures in sequence is prone to generate new emotion meaning that is not captured by categorization of the individual transient stimuli. Cre- ation of new meaning by image juxtaposition was intensively studied in cinema theory 4. Discussion and Conclusion 59 and this is referred as the “Kuleshov” effect (Kuleshov, 1975). A common criticism of studies using pictorial stimuli to induce emotion is grounded in the differentiation between emotion induction and emotion recognition. Some authors have pointed out that subjective ratings of affective pictures can be biased by interpretation of the emotion content of the image. If so, affective ratings do not nec- essarily reflect the affective state of the subject, but the recognition of the emotion that the image is assumed to portrait (Buchanan, Bibas, & Adolphs, 2010). Nonethe- less, normalized affective picture stimuli, such as IAPS, are widely used in emotion research, and subjective ratings of IAPS pictures have been shown to correlate with measures of emotion other then self-assessments, i.e. physiology (review in Cacioppo, John; Tassinary, 2007).

4.5 Future directions

Finally, I will propose a future study inspired by the findings of the work developed in the course of this master. Afterward, I will address aspects of the development of non-invasive therapies based on emotion induction, and how a detailed knowledge of patients affective profile might contribute to the design of such approaches. Following the hypothesis stated above concerning the relation between the potentiation effect found in the study and the cortical non-habituation to repeated stimuli that characterizes the inter-attack phase of migraine, I propose a future study to be conducted along the different stages of a migraine episode, i.e. prodrome, attack, postdrome, and inter-attack phase. In each phase of the migraine episode measurements of affective response patterns to repeated stimulation, and cortical habituation to repeated stimuli, would be conducted. If accepted, my hypothesis would predict that (i) in the attack phase the potentiation of negative affective states resulting from repeated stimulation would be attenuated, and resurge again in the inter-attack phase, and (ii) shifts in affective potentiation of negative states between migraine stages would correlate with shifts in cortical habituation. Future directions for the development of novel non-invasive therapeutic approaches to migraine, based on emotion modulation through perceptual stimuli, have been proposed in the literature (Tommaso et al., 2009; Williams & Rhudy, 2009b). However, a preliminary study by Tommaso et al. (2009) found that sequential presentation of images in regular intervals results in increased cortical wave synchronization, expressed by increased EEG synchronization, that is a predisposing factor for migraine crisis. Hence, questioning the viability of therapeutic approaches based on affective modulation by means of affective pictures presentation. The effects of cortical synchronization observed by Tommaso et al. may result from the onset and offset of picture presentation in regular intervals, as flash stimuli, that can be considered similar to the luminance effect of the onset and the offset of pictures in a computer monitor, was found to lead to EEG synchronization in migraine patients (Angelini et al., 2004). I propose that a continuous flow of affective stimulation composed by video, pictures, and sounds in 4. Discussion and Conclusion 60 a “movie like” fashion could overcome the question of cortical synchronization raised by Tommaso et al.. The growing new media field of interactive narratives presents interesting solutions for the creation of films based on an automated montage of audiovisual material that uses ontological strategies or machine learning methods (Aylett & Louchart, 2003). In addition, game engines under intense development in the computer game industry, can provide inspiration for the creation of an effective emotion modulation system applied migraine therapy. Therapies aiming at the emotion modulation of migraine pain using media technologies, as do similar media-based systems recently developed for the treatment of post-traumatic stress disorders (Cosic,´ Popovic,´ Kukolja, Horvat, & Dropuljic,´ 2010), will have to address the following issues: firstly, take into account the dynamic affective profile of the user in the generation of affective stimulation, to better predict his/her responses to the stimulation; secondly, provide an online assessment of affective responses using physiology and subjective ratings; thirdly, compute a real-time comparison between the predictions of stimulation and the observed affective responses, to guide decisions about stimuli composition; fourthly, automatize narrative composition of audiovisual sequences informed by ontological systems to profit from the presentation of meaningful narrative stimuli; fifthly, provide user defined stimulation protocols to adapt to the patients’ specific sensitivities, e.g. a migraineur suffering form an attack with photophobia symptoms may choose to have only auditory affective simulation. The AMDB software presents a first prototype of a system that allows the dynamic continuous composition of multi-modal affective stimulation. Based on the findings from the empirical study, I propose that composition of audiovisual stimuli for emotion modulation of migraine symptoms should take into account two main aspects, that may be specific to migraineurs affective profile. Firstly, migraineurs enhanced potentiation of negative affective states by exposure to repetitive negative and neutral stimulation. Secondly, migraineurs high susceptibility to disrupt previously induced positive affective states by exposure to negative, or even, neutral affective stimulation. The challenge of creating continuos emotion stimulation, may be addressed by an area of interface between science and art. As narrative techniques developed in the several domains of art have been crafted throughout the history of human culture. A central aspect of narrative techniques is the temporal development of dynamic presentation of circumstances that results in the emotion involvement of the recipients. Hence, it is predictable that the construction of narratives to be used in a clinical context will have to be grounded on the understanding of the specific patterns of affective responses to shifts in emotion content, that may be characteristic to the pathology in question. With the present study I hoped to contribute to the understanding of migraineurs affective responses to changes in emotion context, and provide insights for the design of emotion modulation therapy approaches to migraine, that could in the future help migraine sufferers to ease the disease burden trough the use of highly accessible media technologies. LIST OF FIGURES

1.1 International Headache Society features of migraine (from Goadsby, 2009) ...... 3 1.2 Integrative model for the sequence of events leading to migraine (adapted from Charles & Brennan, 2010 with added information from Alstadhaug, 2009; Burstein & Jakubowski, 2009). This adaptation of the Charles and Brennan model represents migraine as a feedback loop, where migraine symptoms of sensory sensitivity and mood disturbances project back on the action of extrinsic triggers and modulating factors. The model attempts to encompass the heterogeneous clinical presentation of migraine, by proposing that transitions in the sequence may have discrete thresholds, and connections to distinct molecular, cellular, and neurochemical pathways...... 8 1.3 Diagram of structures involved in pathophysiology of migraine, including trigeminovascular nociceptive input transmission and corresponding modulation of that input. Vg = trigeminal ganglion, TCC = trigeminocervical complex, DRG= dorsal root ganglia, PAG = midbrain periaqueductal gray, LC = pontine locus coeruleus, RVM =nucleus raphe magnus (Figure from Goadsby, 2009) ...... 13 1.4 Hypothesized parasympathetic pathway leading to the activation of meningeal nociceptors. SSN = superior salivatory nucleus, SPG = sphenopalatine ganglion, BNST = bed nucleus stria terminalis, LH = lateral hypothalamus, PAG = periaqueductal gray, Pir = piriform cortex, PVN = paraventricular hypothalamic nucleus. (A) The SSN receives input from over 50 limbic and hypothalamic brain areas (red dots). Activity from these areas is assumed to be inﬂuenced by common psychological migraine triggers. (B) Representation of SSN afferents hypothesized to be involved in migraine triggering by olfactory stimuli (Pir), food and sleep deprivation (LH), stress or post stress (PVN, BNST, PAG) (Figure from Burstein & Jakubowski, 2009) . . . 17 List of Figures 62

1.5 Hypothesized mechanism for the genesis of depression symptoms commonly associated with migraine headache by ascending trigeminovascular pathways to the brainstem, hypothalamus and basal ganglia. SpV = spinal trigeminal nucleus, LH = lateral hypothalamus, PAG = periaqueductal gray, PVN = paraventricular hypothalamic nucleus, VP/SI = ventral pallidum/substantia innominata. (A) Trigeminovascular neurons located in the spinal trigeminal nucleus (SpV) project to various limbic and hypothalamic brain areas (represented in the diagram by red dots) whose activity is assumed to underlie common migraine symptoms. (B) SpV projections proposed to be involved in stress (PVN), decreased motivational state (VP/SI), pursuit of solitude (PAG), sleepiness, irritability and loss of appetite (LH) (Figure from Burstein & Jakubowski, 2009) ...... 18

2.1 Graphical representation of the experimental design. The design consisted of two phases: “Single picture phase” and “Transient pictures phase”...... 31 2.2 Schematic representation of the affective picture classes created for this experiment. Horizontal axis corresponds to valence, vertical axis corresponds to arousal. HAU = high-arousal-unpleasant, HAP = high- arousal-pleasant, LAP = low-arousal-pleasant, LAU = low-arousal- unpleasant...... 32 2.3 Transient classes created for this experiment. Horizontal axis correspondes to valence, vertical axis corresponds to arousal. Two affective pictures presented sequential separated by an interval of variable duration (1,3 or 6 seconds). HAU= high-arousal-unpleasant, HAP = high- arousal-pleasant, LAP = low-arousal-pleasant, LAU = low-arousal- unpleasant...... 33 2.4 Graphical user interface of the AMDB ...... 35 2.5 Design for the architecture of the AMDB prototype. Scheme depicts information ﬂow. Green arrows indicate information input, whereas blue arrows indicate information output ...... 36 2.6 Scatterplot of the ratings distribution of the video clips in the valence- arousal space ...... 37

3.1 Differentiation between picture classes for the Control group and the MA group. a) Participant groups are Control (Ctr) and migraineurs with aura (MA). Valence types are Pleasant (P), Neutral (N) and Un- pleasant(U). b) Participant groups are Control (Ctr) and migraineurs with aura (MA). Arousal types are High (H), Neutral (N) and Low(L). Both groups successfully differentiated between arousal classes and valence classes, supporting the affective picture classiﬁcation used in this experiment...... 40 List of Figures 63

3.2 Accuracy of IAPS valence induction predictions for Control and MA group. a) Scatterplot assessing the relationship between mean valence ratings from the Control group and IAPS valence scores. b) Scatterplot assessing the relationship between mean valence ratings from the MA group and IAPS valence scores. IAPS values significantly predicted valence ratings on both groups. Increases of IAPS valence values were always correlated with increases in valence ratings...... 42 3.3 Accuracy of IAPS arousal induction predictions for Control and MA group. a) Scatterplot assessing the relationship between mean arousal ratings from the Control group and IAPS arousal scores. b) Scatter- plot assessing the relationship between mean arousal ratings from the MA group and IAPS arousal scores. IAPS values significantly predicted arousal ratings on both groups. Nonetheless, the linear regression model computed for the MA group performed worse in predicting variances in the arousal scores...... 43 3.4 Valence affective responses to picture types for Control and MA group. Picture classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal- Unpleasant (LAU). Pictures of class LAU were rated by migraineurs as more unpleasant...... 43 3.5 Arousal affective responses to picture types for Control and MA group. Picture classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal- Unpleasant (LAU). Unpleasant stimuli were considered significantly more arousing by migraineurs then by non-migraineurs...... 44 3.6 Relationship between order of presentation and valence scores for Neu- tral pictures. a) Presents the scatterplot for the Control group, where presentation order was found not to be significantly correlated to the affective experience of the neutral content. b) Presents the scatterplot for the MA group, where Neutral pictures’ presentation order and valence ratings were negatively correlated. Denoting that the more neutral pictures participants of the MA group viewed the less pleasant they rated their affective experience of those pictures...... 45 3.7 Relationship between order of presentation and valence scores for Un- pleasant pictures. a) Presents the scatterplot for the Control group, where the presentation order was found not to be significantly correlated to the affective experience of unpleasant stimuli. b) Presents the scatterplot for the MA group, where Unpleasant pictures’ presentation order and valence ratings were negatively correlated. Revealing that the more unpleasant pictures participants of the MA group viewed the more unpleasant they rated their subjective experience of those pictures...... 46 List of Figures 64

3.8 Valence Inhibition effect for Pleasant to Unpleasant transient stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal- Unpleasant (LAU). Valence inhibition amplitudes of transitions starting from a pleasant context were found to be significantly different between groups [Pleasant-HAU: r=2.40, p<0.05; Pleasant – LAU: r=3.10, p<0.01; Pleasant – Neutral: r=2.37, p<0.05]...... 48 3.9 Arousal modulation accompanying the valence inhibition effect for Pleasant to Unpleasant stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal- Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). Significantly different arousal modulation between groups was found only for transitions from Positive to HAU [r=-2.54, p<0.05]...... 49 3.10 Relationship between the arousal of the inhibitor picture and the valence inhibition effect for Pleasant-To-Unpleasant stimuli. a) For the Control group a significant correlation was found, explaining 30% of the valence inhibition data variation [Control : b = -0.99, t(378) = 1.908, p < 0.001, adjusted R2= 0.3 , F(1, 378) = 124.3]. b) For the MA group, a significant correlation found explaining 10% of the valence inhibition data variation [MA: b = -0.49, t(378) = 2.64, p < 0.001, adjusted R2= 0.1 , F(1, 378) = 22.69, p < 0.001]...... 50 3.11 Valence Inhibition effect for Unpleasant to Pleasant transient stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal-Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low-Arousal-Pleasant (LAP) and Low-Arousal- Unpleasant (LAU). Valence inhibition amplitudes of transitions starting from a unpleasant context stimuli were found to be similar between both groups...... 51 3.12 Arousal modulation accompanying the valence inhibition effect for Unpleasant to Pleasant transient stimuli. Groups are Control (Ctr) and migraineurs with aura (MA), inhibitor classes are High-Arousal- Pleasant (HAP), High-Arousal-Unpleasant (HAU), Neutral (Neut), Low- Arousal-Pleasant (LAP) and Low-Arousal-Unpleasant (LAU). Arousal modulation found in Unpleasant-To-Pleasant stimuli was similar between both groups...... 52 List of Figures 65

3.13 Relationship between the arousal of the inhibitor picture and the valence inhibition effect Unpleasant-To-Pleasant transitions. a) For the Control group, a signiﬁcant correlation was found explaining 10% of the valence inhibition data variation [b = -0.49, t(378) = 2.64, p < 0.001, adjusted R2= 0.1 , F(1, 378) = 22.69, p < 0.001]. b) For the MA group a signiﬁcant correlation was found, explaining 2% of the valence inhibition data variation [b = -0.28, t(378) = -2.82, p < 0.001, adjusted R2= 0.02 , F(1, 378) = 7.95, p < 0.001]...... 52 3.14 Valence Inhibition in transitions with different intervals. White bars corresponded to the Control group (1, 2 and 3 seconds correspond to Ctr.1, Ctr.3 and Ctr.6). Filled bars corresponded the MA group (1, 2 and 3 seconds are MA.1, MA.3 and MA.6). Vertical axis indicates the amplitude of the valence inhibition effect...... 53 3.15 Arousal modulation in transitions with different intervals. White corresponded to the Control group (1, 2 and 3 seconds correspond to Ctr.1, Ctr.3 and Ctr.6). Filled bars corresponded the MA group (1, 2 and 3 seconds are MA.1, MA.3 and MA.6). Vertical axis indicates the arousal modulation...... 54 LIST OF TABLES

1.1 Population based studies focusing on the relationship between personality traits and migraine. MO = Migraine without aura, MA = Migraine with Aura, CM = Chronic migraine, EM = Episodic migraine, TTH = Tension-Type Headache...... 5 1.2 Population based studies of migraine comorbidity with psychiatric disorders. Table adapted from Radat & Swendsen, 2005. DSM = diagnostic and statistic manual, DIS = diagnostic interview schedule, CIDI = composite international diagnostic interview, SPIKE = structured psychopathological interview and rating of the social consequences for epidemiology, TTH = tension-type headache...... 7 1.3 Neuroanatomical structures involved in the processing of vascular head pain. Table from Goadsby, 2009 ...... 12

3.1 Mean scores for valence and arousal picture classes...... 40 3.2 Valence and arousal mean scores for each picture class ...... 42 REFERENCES

Abbate-Daga, G., Fassino, S., Lo Giudice, R., Rainero, I., Gramaglia, C., Marech, L., et al. (2007, January). Anger, depression and personality dimensions in patients with migraine without aura. Psychotherapy and psychosomat- ics, 76(2), 122–8. Available from http://www.ncbi.nlm.nih.gov/ pubmed/17230053 Alstadhaug, K. B. (2009, August). Migraine and the hypothalamus. Cephalalgia : an international journal of headache, 29(8), 809–17. Available from http:// www.ncbi.nlm.nih.gov/pubmed/19604254 Ambrosini, A., Magis, D., & Schoenen, J. (2010). Migraine – clinical neurophysiology. Handbook of Clinical Neurology, 18. Andrasik, F., & Rime, C. (2007, May). Can behavioural therapy influence neuro- modulation? Neurological sciences : official journal of the Italian Neurolog- ical Society and of the Italian Society of Clinical Neurophysiology, 28 Suppl 2, S124–9. Available from http://www.ncbi.nlm.nih.gov/pubmed/ 17508158 Angelini, L., Tommaso, M. D., Guido, M., Hu, K., Ivanov, P. C., Marinazzo, D., et al. (2004, July). Steady-State Visual Evoked Potentials and Phase Synchronization in Migraine Patients. Physical Review Letters, 93(3). Aurora, S. K., Barrodale, P. M., Tipton, R. L., & Khodavirdi, A. (2007). Brain- stem dysfunction in chronic migraine as evidenced by neurophysiological and positron emission tomography studies. Headache, 47(7), 996–1003; discussion 1004–7. Aurora, S. K., & Nagesh, V. (2010). Pathophysiology of migraine. In M. A. Moskowitz & G. Nappi (Eds.), Handbook of clinical neurology, vol.97, headache (3rd series ed., pp. 267–273). Elsevier. Aylett, R., & Louchart, S. (2003, December). Towards a narrative theory of virtual reality. Virtual Reality, 7(1), 2–9. Bag, B., Hacihasanoglu, R., & Tufekci, F. G. (2005, May). Examination of anxiety, hostility and psychiatric disorders in patients with migraine and tension-type headache. International journal of clinical practice, 59(5), 515–21. Available from http://www.ncbi.nlm.nih.gov/pubmed/15857345 Bahra, A., Matharu, M., Buchel, C., Frackowiak, R., & Goadsby, P. (2001, March). Brainstem activation specific to migraine headache. The Lancet, 357(9261), 1016–1017. Benoit, D. P. (2009). An introduction to migraine pathophysiology. Techniques References 68

in Regional Anesthesia and Pain Management, 13(1), 5–8. Available from http://dx.doi.org/10.1053/j.trap.2009.04.002 Breslau, N. (1991, April). Migraine, psychiatric disorders, and suicide attempts: An epidemiologic study of young adults. Psychiatry Research, 37(1), 11–23. Breslau, N., & Andreski, P. (1995, July). Migraine, Personality, and Psychiatric Co- morbidity. Headache: The Journal of Head and Face Pain, 35(7), 382–386. Breslau, N., Merikangas, K., & Bowden, C. L. (1994, October). Comorbidity of migraine and major affective disorders. Neurology, 44(10 Suppl 7), S17–22. Breslau, N., & Rasmussen, B. K. (2001). The impact of migraine: Epidemiology, risk factors, and co-morbidities. Neurology, 56(90001), 4S–12. Breslau, N., Schultz, L. R., Stewart, W. F., Lipton, R. B., Lucia, V. C., & Welch, K. M. A. (2000). Headache and major depression: Is the association specific to migraine? Neurology, 54(2), 308–. Buchanan, T. W., Bibas, D., & Adolphs, R. (2010, May). Associations between Feel- ing and Judging the Emotions of Happiness and Fear: Findings from a Large- Scale Field Experiment. PLoS ONE, 5(5), e10640. Burstein, R. (2001, January). Deconstructing migraine headache into peripheral and central sensitization. Pain, 89(2-3), 107–10. Burstein, R., & Jakubowski, M. (2009, May). Neural substrate of depression during migraine. Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology, 30 Suppl 1(0), S27–31. Burstein, R., Yarnitsky, D., Goor-Aryeh, I., Ransil, B. J., & Bajwa, Z. H. (2000, May). An association between migraine and cutaneous allodynia. Annals of neurology, 47(5), 614–24. Buse, D. C., & Andrasik, F. (2009, May). Behavioral medicine for migraine. Neuro- logic clinics, 27(2), 445–65. Available from http://www.ncbi.nlm.nih .gov/pubmed/19289225 Cacioppo, John; Tassinary, L. G. B. G. G. (2007). The Handbook Of Psychophysiol- ogy. Cady, R. K., Schreiber, C. P., & Farmer, K. U. (2004, May). Understanding the patient with migraine: the evolution from episodic headache to chronic neurologic disease. A proposed classification of patients with headache. Headache, 44(5), 426–35. Cao, Y. (1999, May). Functional MRI-BOLD of Visually Triggered Headache in Patients With Migraine. Archives of Neurology, 56(5), 548–554. Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS Scales. Journal of Personality and Social Psychology, 67(2), 319–333. Avail- able from http://doi.apa.org/getdoi.cfm?doi=10.1037/0022 -3514.67.2.319 Charles, A., & Brennan, K. C. (2010). The neurobiology of migraine. In References 69

M. A. Moskowitz & G. Nappi (Eds.), Handbook of clinical neurology, vol.97, headache (3rd series ed., pp. 99–108). Elsevier. Compton, R. J. (2003, June). The Interface Between Emotion and Attention: A Re- view of Evidence from Psychology and Neuroscience. Behavioral and Cognitive Neuroscience Reviews, 2(2), 115–129. Coppola, G., Vandenheede, M., Di Clemente, L., Ambrosini, A., Fumal, A., De Pasqua, V., et al. (2005, January). Somatosensory evoked high-frequency os- cillations reﬂecting thalamo-cortical activity are decreased in migraine patients between attacks. Brain : a journal of neurology, 128(Pt 1), 98–103. Corr, P. J., & Matthews, G. (2009). The Cambridge Handbook of Personality Psychol- ogy (Cambridge Handbooks in Psychology). Cambridge University Press. Cortelli, P., Cevoli, S., Bonavina, G., Magniﬁco, F., & Pierangeli, G. (2001). Do changes of ANS in migraine subjects play a pathogenetic role? Cephalalgia, 73–77. Cosic,´ K., Popovic,´ S., Kukolja, D., Horvat, M., & Dropuljic,´ B. (2010, February). Physiology-driven adaptive virtual reality stimulation for prevention and treatment of stress related disorders. Cyberpsychology, behavior and social network- ing, 13(1), 73–8. Craig, A. D. (2002, August). How do you feel? Interoception: the sense of the physiological condition of the body. Nature reviews. Neuroscience, 3(8), 655– 66. Critchley, H. D. (2005, December). Neural mechanisms of autonomic, affective, and cognitive integration. The Journal of comparative neurology, 493(1), 154–66. Damasio, A. (2001, October). Fundamental feelings. Nature, 413(6858), 781. Damásio, A. (2003, October). Feelings of Emotion and the Self. Annals of the New York Academy of Sciences, 1001(1), 253–261. Davidson, R. J. (1998, May). Affective Style and Affective Disorders: Perspectives from Affective Neuroscience. Cognition & Emotion, 12(3), 307–330. Davidson, R. J. (2000). Affective style, psychopathology, and resilience: Brain mechanisms and plasticity. American Psychologist, 55(11), 1196–1214. Davidson, R. J. (2003). Affective neuroscience and psychophysiology: Toward a synthesis. Psychophysiology, 40, 655–665. Denuelle, M., Fabre, N., Payoux, P., Chollet, F., & Geraud, G. (2007). Hypothalamic activation in spontaneous migraine attacks. Headache, 47(10), 1418–26. Diagnostic and Statistical Manual of Mental Disorders DSM-IV-TR Fourth Edition (Text Revision) (4th editio ed.). (2000). American Psychiatric Publishing, Inc. Ekman, P. (2009, December). Darwin’s contributions to our understanding of emotional expressions. Philosophical transactions of the Royal Society of London. Series B, Biological sciences, 364(1535), 3449–51. Fan, A. Y., Gu, R.-J., & Zhou, A.-N. (1999, September). MMPI Changes Associ- ated With Therapeutic Intervention: A Migraine Control Study. Headache: The Journal of Head and Face Pain, 39(8), 581–585. Fields, R. D. (2009). The Other Brain: From Dementia to Schizophrenia, How New References 70

Discoveries about the Brain Are Revolutionizing Medicine and Science. Simon & Schuster. Frediani, F., & Villani, V. (2007, May). Migraine and depression. Neurological sciences : official journal of the Italian Neurological Society and of the Italian Society of Clinical Neurophysiology, 28 Suppl 2(0), S161–5. Géraud, G., Denuelle, M., Fabre, N., Payoux, P., & Chollet, F. (2005, July). [Positron emission tomographic studies of migraine]. Revue neurologique, 161(6-7), 666– 70. Giffin, N. J., Ruggiero, L., Lipton, R. B., Silberstein, S. D., Tvedskov, J. F., Olesen, J., et al. (2003, March). Premonitory symptoms in migraine: an electronic diary study. Neurology, 60(6), 935–40. Goadsby, P. J. (2009). Pathophysiolo gy of Migraine. Genetic Epidemiology, 27, 335–360. Goadsby, P. J., Lipton, R. B., & Ferrari, M. D. (2002). Migraine - Current Under- standing and treatment. New English Journal Of Medicine, 346(4), 257–270. Haan, J., Kors, E. E., Vanmolkot, K. R. J., van Den Maagdenberg, A. M. J. M., Frants, R. R., & Ferrari, M. D. (2005, June). Migraine genetics: an update. Current pain and headache reports, 9(3), 213–20. Hagemann, D. (2003, June). Central and autonomic nervous system integration in emotion. Brain and Cognition, 52(1), 79–87. Jayaro, C., de La Vega, I., Díaz-Marsá, M., Montes, A., & Carrasco, J. L. (2008). [The use of the International Affective Picture System for the study of affective dysregulation in mental disorders]. Actas españolas de psiquiatría, 36(3), 177– 82. John, P. J., Sharma, N., Sharma, C. M., & Kankane, A. (2007, May). Effectiveness of yoga therapy in the treatment of migraine without aura: a randomized controlled trial. Headache, 47(5), 654–61. Available from http://www.ncbi.nlm .nih.gov/pubmed/17501846 Karakurum, B., Soylu, O., Karata¸s,M., Giray, S., Tan, M., Arlier, Z., et al. (2004, November). Personality, depression, and anxiety as risk factors for chronic migraine. The International journal of neuroscience, 114(11), 1391–9. Available from http://www.ncbi.nlm.nih.gov/pubmed/15636352 Kenntner-Mabiala, R., & Pauli, P. (2005, September). Affective modulation of brain potentials to painful and nonpainful stimuli. Psychophysiology, 42(5), 559–67. Kraft, C. A., Lumley, M. A., D’Souza, P. J., & Dooley, J. A. (2008, February). Emo- tional approach coping and self-efficacy moderate the effects of written emotional disclosure and relaxation training for people with migraine headaches. British journal of health psychology, 13(Pt 1), 67–71. Available from http:// www.ncbi.nlm.nih.gov/pubmed/18230235 Kuleshov, L. V. (1975). Kuleshov on Film: Writings. Univ of California Pr. Lane, R., & Nadel, L. (2002). Cognitive Neuroscience of Emotion (Series in Affective Science). Oxford University Press, USA. Lang, P., Bradley, M., & Cuthbert, B. (2008). International affective picture system References 71

(IAPS): Affective ratings of pictures and instruction manual. Technical Report A-8. Florida. Lang, P. J. (1995, May). The emotion probe. Studies of motivation and attention. The American psychologist, 50(5), 372–85. Lang, P. J., & Davis, M. (2006). Understanding Emotions (Vol. 156). Elsevier. Lawler, S. P., & Cameron, L. D. (2006, August). A randomized, controlled trial of massage therapy as a treatment for migraine. Annals of behavioral medicine : a publication of the Society of Behavioral Medicine, 32(1), 50–9. Available from http://www.ncbi.nlm.nih.gov/pubmed/16827629 Leao, A. A. P., & Morison, R. S. (1945). Propagation of spreading cortical depression. J Neurophysiol, 8(1), 33–45. Levy, D., Strassman, A. M., & Burnstein, R. (2009, June). A critical view on the role of migraine triggers in the genesis of migraine pain. Headache, 49(6), 953–7. Lipton, R. B., Hamelsky, S. W., Kolodner, K. B., Steiner, T. J., & Stewart, W. F. (2000). Migraine, quality of life, and depression: A population-based case-control study. Neurology, 55(5), 629–635. Loeser, J. D., & Rolf-Detlef, T. (2008). The Kyoto protocol of IASP Basic Pain Terminology. Pain, 137, 473–477. Merikangas, K. (1993a, June). Headache and personality: Results of a community sample of young adults. Journal of Psychiatric Research, 27(2), 187– 196. Available from http://dx.doi.org/10.1016/0022-3956(93) 90007-O Merikangas, K. (1993b, June). Headache and personality: Results of a community sample of young adults. Journal of Psychiatric Research, 27(2), 187–196. Merikangas, K. R. (1994, September). Psychopathology and Headache Syndromes in the Community. Headache: The Journal of Head and Face Pain, 34(8), S17– S22. Millan, M. (2002, April). Descending control of pain. Progress in Neurobiology, 66(6), 355–474. Moskowitz, M. A., Bolay, H., & Dalkara, T. (2004, February). Deciphering migraine mechanisms: clues from familial hemiplegic migraine genotypes. Annals of neurology, 55(2), 276–80. Moskowitz, M. A., & Buzzi, G. M. (2010). Migraine: general aspects. In Handbook of clinical neurology (pp. 254–266). Nestoriuc, Y., Martin, A., Rief, W., & Andrasik, F. (2008, September). Biofeedback treatment for headache disorders: a comprehensive efﬁcacy review. Applied psychophysiology and biofeedback, 33(3), 125–40. Available from http:// www.ncbi.nlm.nih.gov/pubmed/18726688 Olesen, J., & Steiner, T. J. (2004, June). The international classiﬁcation of headache disorders, 2nd edn (ICDH-II). Journal of Neurology, Neurosurgery & Psychia- try, 75(6), 808–811. Panksepp, J. (1982, September). Toward a general psychobiological theory of emotions. Behavioral and Brain Sciences, 5(03), 407–422. References 72

Peroutka, S. J. (2004). Migraine: A Chronic Sympathetic Nervous System Disorder. System. Pietrobon, D., & Striessnig, J. (2003). Neurobiology of migraine. Neuroscience, 4(May), 386–399. Porreca, F., Ossipov, M. H., & Gebhart, G. F. (2002, June). Chronic pain and medullary descending facilitation. Trends in neurosciences, 25(6), 319–25. Prinz, J. J. (2006). Gut Reactions: A Perceptual Theory of Emotion (Philosophy of Mind). Oxford University Press, USA. Radat, F., & Swendsen, J. (2005, March). Psychiatric comorbidity in migraine: a review. Cephalalgia : an international journal of headache, 25(3), 165–78. Raskin, N. H., Hosobuchi, Y., & Lamb, S. (1987, September). Headache May Arise From Perturbation of Brain. Headache: The Journal of Head and Face Pain, 27(8), 416–420. Rhudy, J. L., & Meagher, M. W. (2001). The role of emotion in pain modulation. Current Opinion in Psychiatry, 14(3), 241–245. Rhudy, J. L., Williams, A. E., McCabe, K. M., Nguyen, M. A. T. V., & Rambo, P. (2005, September). Affective modulation of nociception at spinal and supraspinal levels. Psychophysiology, 42(5), 579–87. Rhudy, J. L., Williams, A. E., Mccabe, K. M., Russell, J. L., & Maynard, L. J. (2008). Emotional control of nociceptive reactions (ECON): Do affective valence and arousal play a role? Pain, 136, 250–261. Roy, M., Piché, M., Chen, J.-I., Peretz, I., & Rainville, P. (2009, December). Cere- bral and spinal modulation of pain by emotions. Proceedings of the National Academy of Sciences of the United States of America, 106(49), 20900–5. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39(6), 1161–1178. Salomone, S., Caraci, F., & Capasso, A. (2009, January). Migraine: an overview. The open neurology journal, 3, 64–71. Available from http://www.ncbi.nlm .nih.gov/pubmed/19888434 Sánchez-Román, S., Téllez-Zenteno, J. F., Phols, F. Zermeño, García-Ramos, G., Velázquez, A., Derry, P., et al. (2007, April). Personality in patients with migraine evaluated with the "Temperament and Character Inventory". The journal of headache and pain, 8(2), 94–104. Schimmack, U. (2007). Eliciting Mixed Feelings with the Paired-Picture Paradigm: A Tribute to Kellogg (1915). Cognition & Emotion, 21(7), 1546–1553. Schoonman, G., Evers, D., Ballieux, B., Geus, E. D., Kloet, E. D., Terwindt, G., et al. (2007). Is stress a trigger factor for migraine? Psychoneuroendocrinology, 32(5), 532–8. Schur, E. A., Noonan, C., Buchwald, D., Goldberg, J., & Afari, N. (2009). A twin study of depression and migraine: evidence for a shared genetic vulnerability. Headache, 49(10), 1493–502. Stam, A. H., Vries, B. de, Janssens, A. C. J. W., Vanmolkot, K. R. J., Aulchenko, References 73

Y. S., Henneman, P., et al. (2010, January). Shared genetic factors in migraine and depression: evidence from a genetic isolate. Neurology, 74(4), 288–94. Stetter, F., & Kupper, S. (2002, March). Autogenic training: a meta-analysis of clinical outcome studies. Applied psychophysiology and biofeedback, 27(1), 45–98. Available from http://www.ncbi.nlm.nih.gov/pubmed/ 12001885 Swartz, K. L. (2000, October). Mental Disorders and the Incidence of Migraine Headaches in a Community Sample: Results From the Baltimore Epidemiologic Catchment Area Follow-up Study. Archives of General Psychiatry, 57(10), 945– 950. Tan, H. J., Suganthi, C., Dhachayani, S., Rizal, A. M. M., & Raymond, A. A. (2007, April). The coexistence of anxiety and depressive personality traits in migraine. Singapore medical journal, 48(4), 307–10. Available from http:// www.ncbi.nlm.nih.gov/pubmed/17384877 Tommaso, M. de, Calabrese, R., Vecchio, E., De Vito Francesco, V., Lancioni, G., & Livrea, P. (2009). Effects of affective pictures on pain sensitivity and cortical responses induced by laser stimuli in healthy subjects and migraine patients. International journal of psychophysiology : official journal of the International Organization of Psychophysiology, 74(2), 139–48. Torelli, P., & D’Amico, D. (2004, October). An updated review of migraine and comorbid psychiatric disorders. Neurological sciences : official journal of the Ital- ian Neurological Society and of the Italian Society of Clinical Neurophysiology, 25 Suppl 3, S234–5. Available from http://www.ncbi.nlm.nih.gov/ pubmed/15549545 Vasudeva, S., Claggett, A. L., Tietjen, G. E., & McGrady, A. V. (2003, March). Biofeedback-assisted relaxation in migraine headache: relationship to cerebral blood flow velocity in the middle cerebral artery. Headache, 43(3), 245–50. Available from http://www.ncbi.nlm.nih.gov/pubmed/ 12603643 Villemure, C., & Bushnell, M. C. (2002). Cognitive modulation of pain: how do attention and emotion influence pain processing? Pain(95), 195–199. Wiech, K., Ploner, M., & Tracey, I. (2008, August). Neurocognitive aspects of pain perception. Trends in cognitive sciences, 12(8), 306–13. Williams, A. E., & Rhudy, J. L. (2009a). Emotional modulation of autonomic responses to painful trigeminal stimulation. International journal of psychophysiology : official journal of the International Organization of Psychophysiology, 71(3), 242–7. Williams, A. E., & Rhudy, J. L. (2009b). Emotional modulation of autonomic responses to painful trigeminal stimulation. International journal of psychophysiology : official journal of the International Organization of Psychophysi- ology, 71(3), 242–7. Available from http://dx.doi.org/10.1016/ j.ijpsycho.2008.10.004 Williams, A. E., & Rhudy, J. L. (2009c, May). Supraspinal modulation of trigeminal References 74

nociception and pain. Headache, 49(5), 704–20. Wober, C., & Wober-Bingol, c. (2010). Triggers of migraine and tension-type headache. In M. A. Moskowitz & G. Nappi (Eds.), Handbook of clinical neurology, vol.97, headache (3rd series ed., pp. 162–172). Elsevier. Wolf, S. (1963). A New View of Disease. JAMA, 184(2), 129–130. Woods, R. P., Iacoboni, M., & John, C. M. (1994). Bilateral Spreading Cerebral Hy- poperfusion during Spontaneous Migraine Headache. The New England Journal of Medicine, 331, 1689–1692. APPENDIX A. TREATMENT APPROACHES FOLLOWED BY MA PARTICIPANTS

Treatment Number of patients

Pharmacological

Tricycle antidepressants 4 Amitriptilina 10 mg or 25 mg per 24 hours Combination of tricycle antidepressants and anticonvulsants 2 Amitriptilina 10 mg and Topiramato 100mg per 24 hours Combination of anticonvulsants and other drugs 1 Topiramato 75 mg and Lorazepam 1 mg per 24 hours Combination of tricycle antidepressants and other drugs 1 Amitriptilina 10 mg and Agomelatine 25 mg and Telmisartan 20 mg per 24 hours Combination of tricycle antidepressants and beta‐blockers 2 Amitriptilina 10 mg and Nebivolol 2.5 mg or Propanolol 20 mg per 24 hours Combination of ISRS and beta‐blockers 1 Trazodona 50 mg and Nadolol 40 mg per 24 hours Other drugs 1 Betahistina 8 mg per 24 hours

Nonpharmacological

Music therapy 1

No treatment

B. LEGAL CONSENT

You are about to participate in an experiment of the Hospital Vall d’Hebron in collaboration with the Laboratory for Synthetic Per- ceptive, Emotive and Cognitive Systems of the Universitat Pompeu Fabra. The aim of this experiment is to study affective responses to pictures of people who suffer from migraines.

The experimental procedure should take around 40-60 minutes. A set of pictures with different emotional content is to be presented to you. Some of this pictures might have strong emotional impact while others might seem relatively neutral. You will be instructed to give us some feedback about how you are feeling. During the picture presentation we will also record physiological signals. Afterwards we will ask you to fill a questionnaire about how you behave/feel in specific situations.

As a participant you have the right to decline participation and withdraw from the experiment at any time without further consequences. Data recorded will be analyzed and used for developmental and publication purposes, with complete respect for anonymity and confidentiality.

Please do not hesitate to ask questions before or after the experiment. For any further doubts after the experiment please do not hesitate to contact us.

I Hereby confirm my consent to participate in this experiment.

(signature)

Barcelona, 2010 C. BEHAVIORAL INHIBITION BEHAVIORAL ACTIVATION QUESTIONNAIRE (BISBAS) ID:______

En las siguientes preguntas, indique si usted está de acuerdo o en desacuerdo. Por favor responda cada pregunta independientemente una de la otra.

es muy es un es un es muy cierto en poco poco falso en mi caso cierto en falso en mi caso mi caso mi caso 1. La familia de uno es lo mas importante de la vida. 2. Ni siquiera cuando algo malo está por sucederme, yo siento miedo o nerviosismo. 3. Yo hago lo imposible por conseguir lo que quiero. 4. Cuando soy buena en algo que estoy haciendo, me gusta seguirlo haciendo. 5. Siempre estoy dispuesto(a) a tratar algo nuevo si me parece que sería divertido. 6. La manera en que yo me visto es muy importante para mi. 7. Cuando consigo algo que quiero, me siento emocionado(a) y con energía. 8. Las criticas o el regaño me duelen bastante. 9. Cuando quiero algo, casi siempre hago lo imposible para obtenerlo. 10. Muchas veces hago las cosas simplemente por el hecho que serían divertidas en hacer. 11. Se me hace dificil encontrar el tiempo de hacer cosas, como cortarme el pelo. 12. Si veo la oportunidad de conseguir algo que quiero, la aprovecho inmediatament. 13. Me preocupo y me siento mal cuando pienso o creo que alguien esta enojado(a) conmigo. 14. Cuando veo la oportunidad de algo que me gusta, me emociono inmediatamente. 15. A veces, actuo impulsivamente. 16. Si creo que algo desagradable va a suceder, casi siempre me desespero. 17. A menudo pienso en la razón por la cual la gente actua de la manera que lo hacen. 18. Me afecto profundamente cuando me suceden cosas buenas. 19. Me siento intranquilo(a) cuando pienso que no he hecho un buen trabajo. 20. Deseo estimulos y nuevas sensaciones. 21. Cuando estoy en busqueda de algo, para mi no existen barreras. 22. En comparación con mis amistades, tengo pocos miedos. 23. Me emocionaría ganar un concurso. 24. Me preocupa cometer errores.